test: comprehensive scheduler test suite (96 tests)

[aterrylu]

Summary

• Add 96 tests across 3 new test files covering the entire scheduler feature that shipped in PR feat: native cron scheduler with Croner engine, MCP tools, and dashboard UI #132
• Add test utility exports (_resetForTesting, _setExecutors, _onRunCompleted) to scheduler.ts for clean executor mocking
• No functional changes to scheduler behavior

Test Coverage

Test File	Tests	Covers
schedules-crud.test.ts	35	CRUD storage, name validation, run history JSONL append/prune
scheduler.test.ts	27	Engine init/stop, overlap policies (skip/allow/queue), concurrency queue + drain, isolated exec dispatch, agent exec dispatch, one-time schedules, catch-up, state management
schedules-api.test.ts	34	All 9 REST endpoints — validation (bad cron, bad target, missing fields, unsupported policies), happy paths, status/settings

What was untested before

The scheduler shipped with only 3 superficial tests (tool name existence checks). This PR adds coverage for:

• Both execution modes (isolated claude -p + agent gateway send) — verified via injectable mock executors
• Overlap policies — skip blocks when run active, allow passes through, queue/cancel return unsupported error
• Concurrency limits — queuing at max capacity, queue draining on completion
• Run state lifecycle — currentRunId set/cleared, consecutiveFailures increment/reset, runCount tracking
• One-time schedules — future timer setup, past catch-up, invalid date handling
• REST API validation — all required fields, target format, cron expression, overlap policy, duplicate names

Test plan

• All 96 new tests pass
• All 181 total tests pass (existing + new)
• TypeScript type check clean
• Biome lint/format clean

🤖 Generated with Claude Code

[nox-0x]

LGTM. 96 tests covering scheduler engine (overlap policies, concurrency, execution dispatch, state management), REST API endpoints, and CRUD/run-history operations — exactly the coverage this feature needed.