[zaxovaiko]

Summary

• Adds StrykerJS to packages/core, scoped to src/wallet, run against the unit and integration tiers together (vitest.mutation.config.ts): pnpm --filter @openora/core test:mutation.
• New Mutation workflow runs it nightly on dev (and on demand). Incremental: the cached result means only mutants in changed code or tests are re-tested. Uploads the HTML report as an artifact. A report, not a gate.
• real-infra.ts keys the Redis partition on STRYKER_MUTATOR_WORKER too, and OPENORA_TEST_POOL_SIZE caps vitest workers, so parallel Stryker runners don't share a Redis db. Both default to today's behaviour outside Stryker.

Numbers (local, 12 cores)

• 5309 mutants in wallet. Integration tier: 96 min full run, 358 killed, 44 survived, 562 not covered. 89% of covered mutants killed.
• Unit tier alone: 74% of covered, most wallet code not reached. Both tiers now run in one config, so a mutant only a unit test catches still counts as killed.
• Incremental re-run with no change: about 1 min.
• wallet.service.ts is mostly uncovered by either tier. Surviving examples: the webhook signature length check, the deposit-disabled branch, the daily withdrawal cap window and the rate-limit windows.

Notes

• ProfileService.recordTimezone int test can fail on a same-millisecond timestamp compare. If it flakes in the dry run the nightly run aborts; a separate fix.
• A per-PR run on changed lines only looks affordable (about 5-7 min per 100 testable mutants) but isn't measured yet.