feat(spam): capture moderated comments as training samples (axis-2 L2)

[mashbean]

What

L2 of the spam-data-retention roadmap. Emits de-identified labeled samples to SQS at the moderation boundary, so the spam-model training signal survives content deletion that L1's passive DB extraction can't recover:

• clearCommunityWatchOriginalContent nulls the snapshot,
• account archival/ban can purge content.

A separate Lambda worker (in spam-detection-scaffold/workers/spam_sample_worker.py) consumes the queue and appends to the S3 training bucket.

How

• common/notifications/spamSample.ts — enqueueSpamSample, mirrors enqueueReportAlert: best-effort SQS, never throws, no-op when the queue/salt is unconfigured. Comment/author ids are HMAC-SHA256(salt) at emit, so no raw ids ever enter the queue; only the text the model trains on is carried verbatim.
• Wired:
  • communityWatchRemoveComment → confirmed spam at removal time.
  • clearCommunityWatchOriginalContent → capture before the snapshot is nulled; a reversed action ⇒ hard-negative ham.
• env: MATTERS_AWS_SPAM_SAMPLE_QUEUE_URL, MATTERS_SPAM_SAMPLE_HASH_SALT.

Privacy / governance

De-identified at source (HMAC ids, no contactable PII); only model-relevant text retained. Legal cleared the retention direction (2026-06-13); concrete retention period TBD. The training corpus lives in a separate S3 bucket decoupled from the operational DB so account deletion still fully removes user-facing content.

Off by default

No-op until ops provisions the SQS queue + MATTERS_SPAM_SAMPLE_HASH_SALT and deploys the worker. Zero behavior change otherwise.

Test plan

• With queue+salt set: removing a comment / clearing original content enqueues exactly one de-identified message; ids are hashes, not raw.
• Queue outage / unset → mutation still succeeds (best-effort).
• reversed action via clear path → label=0 (ham).

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 72.59%. Comparing base (3200027) to head (9870c0a).
⚠️ Report is 19 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #4846      +/-   ##
===========================================
- Coverage    73.04%   72.59%   -0.46%     
===========================================
  Files         1066     1067       +1     
  Lines        21659    21178     -481     
  Branches      4816     4623     -193     
===========================================
- Hits         15821    15374     -447     
+ Misses        5357     5326      -31     
+ Partials       481      478       -3     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.