Abort postgres queries whose client died

[WilfordGrimley]

Closes #462

Problem

Postgres backends survive client-container recreation and keep computing for dead clients. Observed 2026-07-25: six SELECT DISTINCT cardpicker_card.id backends ran up to 2h31m after their django/worker client containers were recreated by a deploy — a backend mid-query never touches its client socket, so it can't notice the client is gone until results are ready to send.

Fix

docker/docker-compose.yml postgres service now sets command: postgres -c client_connection_check_interval=30000 (30s) — Postgres 14+ probes the client socket periodically during query execution and aborts the query if the client is gone. docker-compose.prod.yml extends this service and does not override command, so the flag inherits into prod.

Deliberately not statement_timeout (out of scope per the issue) — BULK-mode eligibility scans legitimately run for minutes-hours with a live client and must not be killed.

Verification

• docker compose -f docker/docker-compose.yml config (dev file, no sudo) renders command: [postgres, -c, client_connection_check_interval=30000] for the postgres service.
• docker compose -f docker/docker-compose.prod.yml config (with dummy env vars for interpolation, no sudo) renders the same command inherited via extends, confirming prod doesn't override it.
• Isolated throwaway container (docker run --name pg462check ... postgres -c client_connection_check_interval=30000, no connection to the live prod DB or its volume): SHOW client_connection_check_interval; → 30s. Container removed after the check.
• docs_lint.py clean; pre-commit clean on both changed files.

Deploy step (not done in this PR)

Rolling this out requires one deliberate postgres container recreate (docker compose -f docker-compose.prod.yml up -d postgres or equivalent) — a few seconds of DB downtime. django and worker reconnect on their own; no restart needed for either. Documented in docs/infrastructure.md's deploy section.

Checklist

• docker/docker-compose.yml: added command: postgres -c client_connection_check_interval=30000 to the postgres service
• Verified docker-compose.prod.yml extends without overriding command
• Verified docker compose config renders it (dev + prod)
• Verified a live container answers SHOW client_connection_check_interval; with 30s
• docs/infrastructure.md deploy section note (rollout requires one deliberate postgres recreate)
• statement_timeout untouched (out of scope)