Add preflight checks to weekly_collector entrypoint

[cipher813]

Summary

Addresses the class of failure surfaced 2026-04-14: daily_append silently not writing to ArcticDB for two weekdays because `arcticdb` wasn't in the deploy image and the import error was swallowed. A preflight check catches this in ~1s instead of letting the pipeline run to "success" with stale data.

Pattern D (simplest): inline `_preflight()` in `weekly_collector.py`, called from `main()` after config load, before `run_weekly()`. No new files, no shared library. If the check pattern proves valuable across the other modules, we can extract a common helper later — but the per-repo checks are small enough (~30 LOC) that inlining is fine for now.

Scoped to external-world handshakes (env vars, S3, ArcticDB) — not correctness of the collection itself. The hardened collectors from #24 + #25 still own data-integrity hard-fails.

Checks by mode

Mode	Checks
daily	`AWS_REGION`, S3 bucket reachable, ArcticDB `universe` readable, SPY freshness ≤ 4 days
phase1	`AWS_REGION` + `FRED_API_KEY` + `POLYGON_API_KEY`, S3 bucket reachable
phase2	`AWS_REGION` + `FMP_API_KEY` + `EDGAR_IDENTITY`, S3 bucket reachable

The 4-day SPY staleness threshold covers Fri → Tue long-weekend runs with a 1-day buffer, and would have caught today's bug (ArcticDB last wrote 2026-04-12; preflight on 2026-04-14 would have computed age=2d and passed, on 2026-04-15 age=3d still passes, on 2026-04-16 age=4d still passes, on 2026-04-17 age=5d fires). I picked 4 deliberately — want the buffer to absorb long weekends but not a full week of silent failure. Open to tightening to 2–3 days if false positives on long weekends prove rare.

Failure path

`_preflight` raises `RuntimeError` → caller propagates → `main()` exits non-zero → SSM command exits non-zero (set -eo pipefail from #24) → Step Function `Catch [States.ALL] → HandleFailure` fires. Once #26 is deployed, flow-doctor dispatches email + GitHub issue for the ERROR log.

Out of scope (tracked)

• Same pattern in predictor inference + training, research Lambda, executor entrypoints, backtester. Rolling out after this first consumer proves the shape.

Test plan

• `pytest tests/ --ignore=tests/integration -q` — 41 passed
• Syntax check on modified file
• Next DailyData run (2026-04-15) — confirm `Pre-flight OK (mode=daily)` appears in /var/log/daily-data.log before `COLLECTING: daily closes`
• Forced failure: `FRED_API_KEY= python weekly_collector.py --phase 1 --dry-run` should raise at preflight, not deep in the collector

🤖 Generated with Claude Code

[cipher813]

Superseded by the shared-library approach. Preflight logic moved to alpha-engine-lib v0.1.0 (https://github.com/cipher813/alpha-engine-lib). Follow-up PR in this repo will add alpha-engine-lib as a dep and replace this inline version with DataPreflight(BasePreflight).