docs(run-a-node): document which source answered a wallet balance read

[MichaelTaylor3d]

Docs half of dig_ecosystem#2233 (S1 of the #2232 sync plan). Node half: DIG-Network/dig-node#189. Contract half: DIG-Network/dig-node-control-interface#5.

control.wallet.balance now returns a source field, and its synced / peak_height fields changed meaning: they describe the source that answered, not the node in general. docs/run-a-node/manage.md described the old behaviour, which is a bug per §4.3.

Updated: the result JSON sample, the field table, and a new short "Where the answer came from" subsection written for audience 1 (node operators) — plain language, no protocol jargon, leading with the consequence that matters to a person: a "fallback" answer means their address was sent to a public service.

This also makes an existing claim in point-a-consumer.md — that reading from your own node is more private because "your addresses aren't disclosed to a public operator" — checkable by the reader rather than something they have to take on trust.

No i18n changes needed: peak_height appears only in the English tree (grep -rl peak_height i18n/ is empty), so no locale copy is being left stale.

0.13.0 → 0.14.0 (minor: new documented capability).

git diff --shortstat and --ignore-cr-at-eol agree (17/4), so no CRLF flip.

[MichaelTaylor3d]

VERDICT: PASS — reviewed at 84ec5461fc861848b1765ec6d87855771b0dcc23.

Verified:

• Describes shipped behaviour, not intent. The sample, the field table and the new subsection match dig-node#189's actual arms: "fallback" ⇒ synced:false and peak_height:null, "db" ⇒ the replica's peak. The source row correctly documents null as "a node too old to say", which matches the contract crate's Option tri-state rather than inventing a default.
• §4.3 clean. No reference to SYSTEM.md / SPEC.md / CLAUDE.md / internal tickets; no drift commentary; audience-1 voice throughout. The consequence a person actually cares about — the address you asked about was sent to that service — leads, which is the right ordering for this audience.
• No stale locale copy. I re-derived the claim independently: peak_height appears in exactly one file tree-wide (docs/run-a-node/manage.md), and the 13 localised manage.md files are the shorter, pre-balance-section version (46 lines vs 130) with no balance content at all — so nothing localized goes stale. Their incompleteness is a pre-existing gap, not this PR's.
• The claim in point-a-consumer.md:46 ("your addresses aren't disclosed to a public operator") does become checkable by the reader for the first time, as the body says.
• --shortstat matches --ignore-cr-at-eol (17/4). 0.13.0 → 0.14.0 is right for new documented capability.

No findings. Nothing open from me on this PR.

One cross-PR heads-up, not a change request here: I have raised a gating item on dig-node#189 about the tier tracing event being debug! under an info default filter. It does not affect any sentence on this page — nothing here promises the tier appears in the log.