feat(telemetry): classify laps and model analysis eligibility

[Korupt-virus]

Summary

• Separate structural lap validity from lap phase, race conditions, recording quality, and analysis-specific eligibility.
• Preserve valid out, in, pit, grid-start, caution, slow-zone, and formation laps while excluding unsuitable evidence from normal pace analysis.
• Add versioned telemetry-quality facts and deterministic eligibility policies shared by server, client, reports, AI, experiments, live strategy, and future ML consumers.
• Expose concise quality status, detailed reasons, rebuild state, and raw-evidence retention consequences through API and UI.

Change inventory

Lap classification and pace selection

• Model lap phase (flying, out, in, pit, grid start), independent race conditions (caution, slow zone, formation), and pace eligibility as separate persisted fields.
• Detect F1 race-start laps and supported pit/caution states without marking useful telemetry invalid.
• Migrate legacy lap classifications while retaining raw recordings and historical laps.
• Centralize normal-pace selection across Compare, best-lap and recap metrics, stints, driver profiles, experiments, leaderboards, and live strategy.
• Render consistent lap-status labels, tones, and explanations in session, track, comparison, and analysis surfaces.

Telemetry quality and evidence modeling

• Add versioned contracts for recording lifecycle, participant/source provenance, channel fidelity and freshness, localized quality facts, gap summaries, and eligibility decisions.
• Keep stable machine-readable reason codes with severity, affected time/distance ranges, semantic channels, event IDs, and user-readable explanations.
• Measure expected/observed frames, missing count/fraction, largest contiguous gaps, ordering faults, reconnects, timeline discontinuities, lap-boundary confidence, track/position coverage, incidents, and source limitations.
• Preserve player/opponent differences and direct, derived, simplified, stale, pit-only, or unavailable channel evidence without replacing missing values with zero.

Shared eligibility policies

Version 1 includes:

• official timing
• normal pace
• lap comparison
• corner trace
• transient event
• fuel burn
• tire analysis
• stint/falloff
• setup analysis
• driver profile
• ML training

Policies return eligible, eligible with warning, ineligible, or unknown plus policy version, reason codes, supporting evidence, required channels, and optional affected range. Broad analyses can tolerate minor localized gaps while strict transient analysis requires stronger cadence and contiguous coverage.

Persistence, APIs, consumers, and retention

• Persist quality summaries, eligibility sets, schema/policy/config versions, source/output generations, and durable event links.
• Keep live, replay, rebuild, iRacing IBT, MoTeC, raw, canonical-archive, remote-collector, and opponent evidence on explicit source contracts.
• Add lap/session quality endpoints, rebuild status/actions, quality-update notifications, and evidence-retention assessment before raw cleanup.
• Gate comparison, corner analysis, setup recommendations, profiles, experiments, reports, AI prompts, and callouts through shared decisions rather than duplicated filters.
• Add localized quality badges and explanations, including safe rebuild actions when source recordings remain available.

Issue alignment

• Classify first race-start lap as invalid starting lap #222: F1 race-start Lap 1 becomes a preserved grid-start lap, excluded from flying-lap comparison instead of being misrepresented as normal pace.
• Add classification for valid non-pace laps #229: Validity and classification are independent; out, in, pit, grid-start, caution, slow-zone, and formation states remain visible while shared pace filters exclude non-pace evidence.
• Add telemetry quality, confidence, and analysis-eligibility modeling #236: Implements separate lifecycle/channel/lap/classification/eligibility layers, localized facts, stable reasons, source and participant provenance, 11 versioned policies, persisted decisions, API/UI exposure, rebuild parity, and downstream consumer integration.
• Related work: Builds on Finalize the authoritative telemetry catalog foundation #205 channel semantics; supplies quality/provenance inputs for Canonical telemetry archive and hierarchical analysis model (stint → lap → corner → segment) #232, Add analysis provenance receipts and safe rebuild tracking for derived session artifacts #233, Add a unified durable race-event timeline #234, and Make stints and pace segments first-class session records #235; exposes retention consequences for Preserve lap times and metadata while removing raw recordings #217; and provides shared evidence guards relevant to Driver profiling / improvement plan (driving-style profile + ranked coaching) #118, Refactor telemetry ingestion for iRacing, LMU, and future sims #130, Separate telemetry collector and processing server for shared-memory games #215, and Add CrewChief-style opponent lap pace announcements #221. Those related issues remain open and are not closed by this PR.

Verification

• bun test test/tooling/changelog.test.ts test/lap-analysis/quality-model.test.ts test/lap-analysis/quality-source-parity.test.ts test/db/quality-event-linkage.test.ts test/routes/quality-routes.test.ts test/session-capture/import-quality-errors.test.ts test/client/lap-quality-badge.test.ts --timeout 60000
• 47 tests passed across 7 files; 0 failed.

Closes #222
Closes #229
Closes #236

[blacksmith-sh]

Found 4 test failures on Blacksmith runners:

Failures

Test	View Logs
src/stories/ dashboards.snapshot.ts:17:3 › snapshot: ComboDash1 ────────────────────────────────	View Logs
src/stories/ dashboards.snapshot.ts:17:3 › snapshot: SetupBrowserReadOnly ──────────────────────	View Logs
src/stories/ experiment-flow.snapshot.ts:100:3 › renders: review (car focus) ───────────────────	View Logs
src/stories/ experiment-flow.snapshot.ts:100:3 › renders: review (driver focus) ────────────────	View Logs

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[github-actions]

🖼️ App UI changes

Current PR screenshots compared with main. Storybook render-vs-committed-baseline failures are listed separately from committed PR-baseline-vs-main changes.

Changed: responsive /desktop /fm23-driver

Before	After	Diff

Changed: responsive /tablet /fm23-driver

Before	After	Diff

_{Rendered from Storybook and responsive app screenshot tests against current main.}