Integrate LINA LIFE with owned memory, persona, context and shared resources

[thisisjun786]

Change

Consolidate LINA LIFE and the LINA-owned memory, persona, context and shared-resource engines into one review candidate targeting dev.

This PR supersedes #3 and #6, which are now closed with their original branches preserved. Merge commit 6970941fb95090087f3e8ebc5d6ea0d7295057d0 preserves both original heads as parents:

• LIFE/world/social/work/image/recovery: Implement LINA LIFE world, social, work and image engine with recovery #3, cefaffcbb4d767848ace6bc0151b9271bf336bf2.
• Owned memory/persona/context/shared resources: Implement LINA-owned memory, persona, context and shared resource engines #6, d75a11f6a617f9de02cd1246ae27d87d94c4cce7.
• Included target baseline: dev at cd74c89ea642735a9eeee9f63f88651b900a7bef.

The three overlapping files retain the already-ported LIFE recovery behavior, personal-growth connection, publication v3 assertions and owned-memory lifecycle tests. All runtime source remains identical to #6. Tests differ only by the explicit integration budget described below; their assertions are preserved. The merge restores #3's LIFE/recovery documentation, prior image QA evidence and synthetic image QA entry point. Original branches are preserved; neither source PR is merged into dev.

Consumer/review entry points: LIFE engine, context-engine acceptance, UI contracts, and integration record.

Verification

Fresh local checks on the integrated source:

• Both source heads and the target baseline pass Git ancestor checks. No unresolved conflicts; runtime blob equivalence to Implement LINA-owned memory, persona, context and shared resource engines #6, preserved test bodies, and image QA script equivalence to Implement LINA LIFE world, social, work and image engine with recovery #3 verified.
• Focused LIFE/session/storage-isolation/publication/persona regressions: 23 passed, 0 failed, 141 assertions across 5 files.
• LINA_LIFE_NATIVE_TEST=1 LINA_AUTHOR_NATIVE_TEST=1 bun test packages/lina-codex: 268 passed, 0 failed, 1,361 assertions across 45 files. Real local subprocesses, synthetic providers; no external model calls.
• bun run typecheck, bun run lint, bun run ci:validate, bun run ci:build passed. Lint retains 26 existing warnings.
• Independent plan audit and integration-only review passed with no blockers. The reviewer's documentation tense clarification is fixed. Hosted run 34328922004 passed changes, types, lint and build, but package tests ended 3,499 pass / 47 skip / 1 fail. The sole failure was the inherited image-event test budget below; dev-gate correctly failed. Current-head hosted results are reported by GitHub checks.

Risks and remaining work

This consolidation creates the baseline for a separate detailed feature review. It does not certify the full feature set, production data migration, installed runtime, UI journeys, long-conversation quality or live narrative/image quality. Historical provider evidence in the source PRs remains scoped to its original revisions and synthetic scenarios.

No dev/main merge, branch deletion, release, deployment, installed-data change or external model call is part of this task. The consolidation is complete; this PR remains Draft for the owner's separate detailed review. #3 and #6 were closed as superseded, and their source branches remain available. The independent preservation review does not replace the deferred detailed feature review.

Explicit image integration test budget

Two previous hosted runs hit the unchanged Bun default of 5 seconds in packages/lina-runtime/test/life-image-fleet-event.test.ts (5,017ms and 5,027ms). The case covers Fleet setup, accepted event, publication, image persistence, HTTP/proxy reads, cold restart and revocation. There is no separately defined five-second product latency requirement.

The final candidate gives this one functional test an explicit 30-second bound, matching the existing life-e2e.test.ts integration convention. All 16 assertions, production code, other tests and CI job limits remain unchanged. The focused case passes locally in 2.98s with 16 assertions; root/browser typecheck and lint pass on the final candidate. This is a test-budget correction, not evidence of faster runtime performance. Historical timeout results remain documented in the integration record for the later detailed review.