`PRHealthScore` inserts a new DB row on every `synchronize` event — unbounded table growth on active

[BHUVANSH855]

Bug Report

Summary

PRHealthWorkflow.score_pr() always inserts a new PRHealthScore row on every call. Because it's triggered on both opened and synchronize PR events, every push to a PR branch creates a new row. A PR with 30 force-pushes accumulates 30 rows. Over time this makes the pr_health_scores table grow without bound and corrupts averages and dashboard stats.

Location

• app/workflows/prhealth.py — score_pr() (~line 48)
• app/github/webhooks.py — _handle_pull_request() (line ~190–196)

Root Cause

# prhealth.py — always inserts, never updates
record = PRHealthScore(
    owner=owner, repo=repo, pr_number=pr_number, ...
)
db.add(record)   # ← new row every time, no upsert

# webhooks.py — fires on opened AND synchronize
if action in ("opened", "synchronize", "reopened"):
    await wf_health.score_pr(ctx, payload)

Impact

• pr_health_scores table grows proportionally to commits × PRs with no cleanup mechanism
• /api/v1/pr-health and /api/v1/repos/stats return inflated counts and incorrect averages (all historical rows for the same PR are averaged together rather than using the latest score)
• Dashboard PR health charts show duplicate entries for the same PR number

Expected Behavior

Re-scoring a PR should update the existing row for that (owner, repo, pr_number), not insert a new one.

Suggested Fix

Upsert instead of insert:

from sqlalchemy.dialects.postgresql import insert as pg_insert

stmt = pg_insert(PRHealthScore).values(
    owner=owner, repo=repo, pr_number=pr_number, ...
).on_conflict_do_update(
    index_elements=["owner", "repo", "pr_number"],
    set_={...}
)
await db.execute(stmt)

Or add a UniqueConstraint("owner", "repo", "pr_number") to the model and use SQLAlchemy merge(). A migration adding this unique constraint should also deduplicate existing rows.