Address accuracy findings and make map claims reviewable

[0xfauzi]

Source extraction and recorded map reviews could disagree with current code. This PR keeps unreadable source visible, makes historical comparisons use the requested inputs, and reopens reviews when their source or claims change.

• Fix Python and TypeScript extraction, import resolution, public API fingerprints, test identities and configuration freshness. Identical checkouts and supported Python versions produce stable evidence.
• Give journey entries exact coverage, reject invalid drafts, preserve existing positional constructors and restore earlier model files after a later write failure.
• Bind card reviews, flow citations and exact judgement answers to current source. Refresh the self-map and document how to review reopened claims.

Exploratory reports, captured datasets and the unfinished ownership-evaluation harness are excluded. The benchmark README retains measured results, failed acceptance bars and links to the original artifacts in Git history. Production code and its regression tests are unchanged by this cleanup.

Validation after cleanup:

• Python 3.11: 483 tests passed, one expected Python 3.12 syntax test skipped.
• Python 3.13: all 484 tests passed.
• Pre-commit and strict mypy passed.
• Map extraction, rendering, check and strict judgement passed: 45 modules mapped and no open items.
• Cross-version syntax experiment: zero hash differences across 45 modules, unchanged existing Python 3.11 hashes, formatting preserved and body edits detected.

The ownership pilot missed its preset false-challenge bar and does not justify a default policy change. Journey write recovery handles reported I/O failures; process termination and concurrent edits remain outside that contract.

[github-actions]

What this change does to the map

56716e947cd94ad3c0b26f2937bbbfbd9b0484fc to 3a2a1fb76d9efa260075cbfb99773012e19cd1c4: 25 modules changed, 5 added, 0 removed, 0 moved; 4 of 16 cards named.

Changed, nothing to do (5)

• added: systemap.card_review, claimed by Model
• added: systemap.delta_report, claimed by ChangeDetector
• added: systemap.journey_coverage, claimed by Judgement
• added: systemap.judgement_evidence, claimed by Judgement
• added: systemap.model_write, claimed by SecondOpinion

Next to the change

ChangeDetector holds most of what changed, and a flow joins it to 10 of the 16 cards, so naming them would say nothing.

The committed map at 3a2a1fb; the change map itself is systemap figure --mode change --base 56716e9 --out change.svg.
Nothing to decide: the map already covers this change. systemap refresh brings the facts up to date.