[devYRPauli]

Problem

A run of the embedded scenarios records its corpus as custom. Only wic-50 v1 is frozen, and the live scenario set no longer matches v1. The site then shows "catalog unavailable" for every row of a fresh run.

This PR is stacked on #4. Merge #4 first.

Fix

• crates/wic-core/catalogs/wic-50-v2.json freezes the live set as wic-50 revision v2. Its hash is sha256:9661db2f98e15d84a5191948478bace511572fc5d6f24ca3102025b28eccdfe5.
• The runner and the site match a corpus hash against every frozen catalog, not only v1.
• A guard test fails when the embedded scenarios differ from v2. A scenario edit therefore needs a new frozen revision.

Testing

• cargo fmt --check and cargo clippy --workspace --all-targets -- -D warnings are clean.
• cargo test --workspace: 154 tests pass. Fix the three Codex findings on #2 #4 has 150.
• site_joins_a_wic_50_v2_result_to_the_frozen_v2_catalog gives the site a result that names wic-50 v2. The test checks for the scenario description and for no "catalog unavailable" marker in appendix.html and results.json.
• Each new test fails under its mutant:
  • The runner from Fix the three Codex findings on #2 #4: the embedded run records custom, not wic-50.
  • select_catalog searches only the first frozen catalog: the later-catalog test fails.
  • load() keeps only the first frozen catalog: the v2 site test fails.
  • One embedded scenario edited: the guard test fails.
• A run recorded on Fix the three Codex findings on #2 #4 gives 200 "catalog unavailable" markers in the site. With this branch, the same result gives 0.
• willitcall site --results results/ gives output that is byte-identical to Fix the three Codex findings on #2 #4.
• The tools unittest and the redaction ledger check pass.

This PR and the Windows model path PR change the same import block in site/data.rs. The second one to merge needs main merged in first. The merged state of both passes all 155 tests.