Add card estimates and derived participant rollups

[Chris0Jeky]

Cards now carry an optional effort estimate, and board readers can inspect current board, column, participant and unassigned totals. Blank remains unknown, explicit zero stays known, and hours/minutes inputs are available in shared card details and Paper/Legacy quick-create. Multiple assignees each receive the full estimate with a visible overlap explanation; parent estimates remain independent.

The complete vertical covers direct writes, reviewed proposals, chat/MCP, audit/provenance, board/account portability and additive SQLite migration/rollback. Intentional estimate changes retain caller version pins and explicit set/clear semantics. Same-valued proposals also perform an atomic persisted-version check before accepting a no-op, preserving unchanged versions and avoiding false successful receipts after competing writes. Ordered same-card operations and planned creates remain supported.

The editor preserves newer drafts while advancing successful-save versions. Quick-create reconciles with a detail payload that arrives before its POST continuation, preventing duplicate cards/counts while preserving fresh-store creation and navigation ownership. The integration also preserves reviewed source PRs #3053 (planning/backup references) and #3055 (permission recovery).

Validation:

• Focused domain/application/API, real SQLite concurrency and migration/rollback, portability and MCP transport/scope tests pass. Independent reviews cover the core contract, portability/MCP, editor, proposal/rollup integration and both final correctness fixes.
• Full backend: 9,727 passed, 34 existing skips and four architecture failures. The bounded declaration/inventory correction passes 28 architecture tests (one existing skip) and four rollup API tests; later hosted backend/API runs passed. The broad local run was not repeated after the bounded corrections.
• Full frontend: 6,856 passed / three existing skips across 440 files; typecheck/build pass. Final panel/permission integration passes 106 focused tests. Hosted Windows/Linux exposed the first reconciliation fix's fresh-store regression; the final fix passes all 50 direct/integration store tests.
• Real API + SQLite + Chromium: both Legacy/Paper estimate journeys pass on the final frontend (53.4 seconds), covering unknown/zero/90 persistence, clear/resave, exact totals and assignee overlap, edit/move/archive/refresh, Viewer access, 390px layout and Escape focus. Two Paper focus setup regressions were reproduced and repaired without weakening focus/J/Enter assertions; both pass (17.9 seconds).
• Same-valued proposal race: two real SQLite competing-writer controls fail with HTTP 200/Applied on the original branch and pass with HTTP 409 after the fix. All 36 focused service/API/persistence tests pass; independent review is clean and reviewed source bytes match the integrated Git blobs.
• Documentation links and GitHub operations governance pass. Source commits and local evidence are preserved.

Exact final-head required hosted CI must pass before merge. Separate bounded follow-ups retain backend-less demo totals (#3056), rollup snapshot consistency (#3058), request timeout/recovery (#3063), and older-reader export protection (#3065). Physical-device/screen-reader acceptance, skipped PostgreSQL/Docker cases and release/hosting decisions remain in OUTSTANDING_TASKS.md.

Closes #2093
Closes #3044
Closes #3049
Closes #3062

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 51e12925c3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Chris0Jeky]

All reviewed source commits are preserved in pushed integration 51e12925c381991c04bc458535d22afd7eefd200. Completed estimate/editor/portability/proposal/rollup/browser and planning/permission source worktrees were removed with plain git worktree remove only after fresh absolute-path, clean-status, ancestry and remote-head checks.

Durable evidence remains under C:/td0912-evidence: original estimate/permission logs, estimate-browser, finished-estimate-source-artifacts (52 files copied and hash-verified), and estimate-source-pre-removal.json. The active integration and stacked archive follow-on stay available. Branches are retained. The estimate-core worktree remains preserved because its earlier local EF-tool cleanup was denied; no alternate deletion path was used. The original OneDrive checkout and short-clone root are clean and unchanged.

[Chris0Jeky]

Hosted required run 34707276360 passed backend, Windows API, frontend, migration/security and other required jobs, then failed two Paper E2E tests at their initial fixed 40-Tab opener setup. Local actual-API Chromium reproduced both failures and recorded the active-element sequence: the new Estimates control pushed the first card beyond that bound. This is an evidenced setup regression, not a flaky classification.

Head 7aba61d changes only the two setup paths and STATUS. Each now focuses the opener, steps back to its preceding control, then enters with an actual Tab; the exact focus-visible outline and J/Enter selected-card assertions remain unchanged. Final narrow run: 2 passed, 17.9 seconds. ESLint for the spec, 700-file relative doc links, docs governance and diff whitespace pass. No unrelated runtime or gate changes.

A fresh required hosted run is owed at this head. Baseline/final logs and traces were copied and hash-verified under C:/td0912-evidence/estimate-focus-gate (18 local files). They contain synthetic session material and remain local, outside Git/public attachments. Playwright-owned servers exited. The archive child #3059 has merged this parent correction without conflict edits and must still follow oldest-base-first delivery.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 958e9947a0

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 66a9f3e83d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Chris0Jeky]

Card-create reconciliation now preserves a board-detail commit that arrives before the POST continuation, instead of appending the same ID twice or increasing its column count again. It also avoids inserting a late result into a different selected board.

The first fix exposed an existing fresh-store integration contract. Final 66a9f3e83dda2171a46d95601bdc50ee690a2990 preserves that contract as well: failing-before direct/integration controls became 28/28 and 22/22 passing. The independent bounded verification resolved its HIGH finding and found no remaining blocker in the two-file diff. The fresh/null-state ambiguity after clearing a store predates this fix; no actionable UI path was established in that review, so it was not expanded into another fix.

Hosted Windows and Linux at the previous head each failed only that same fresh-store test (6,857 passed, one failed, three skips). Their logs are retained locally. Final real-API Paper/Legacy estimate journeys and the new hosted head gate remain pending. Refs #3062.

[Chris0Jeky]

Merged as 54e4c0a after required workflow34713059470 passed at reviewed head28c4f49de3c761b1212a4313d8746488e28918f0. The merge preserves #3053/#3055; GitHub confirms issues2093/3044/3049/3062 closed and the seven delivery items are Done. First post-merge check: zero unresolved threads on #3054/#3053/#3055.

The archive and relation children have been retargeted to main; their changed bases must pass fresh hosted qualification. No release or physical-device/screen-reader acceptance is inferred. Bounded followups3056/3057/3058/3063/3065 and OUTSTANDING_TASKS.md remain open.

[Chris0Jeky]

Post-merge worktree closeout: the finished estimate integration tree was clean, its tip28c4f49de is preserved by verified remote main54e4c0a86, and all18 retained focus-gate logs/traces/SQLite evidence files were copied and SHA-256 checked under C:/td0912-evidence/delivered-estimate-integration. The copied-hashes.json and pre-removal.json receipts identify every survivor. Plain git worktree remove completed; the local source branch remains. Other worktrees with earlier automatic cleanup denials were excluded and remain preserved.