[fix] Store background task timestamps as naive UTC

[theluckystrike]

Fixes #5106

CodeChecker store on the async path fails with a false TaskTimeoutError when the server's PostgreSQL session runs in a non-UTC timezone. The background task columns (enqueued_at, started_at, finished_at, last_seen_at) are timezone-less DateTime, but BackgroundTask writes aware datetime.now(timezone.utc) values. psycopg2 sends those as timestamptz, and PostgreSQL stores them as local wall time. _db_timestamp_to_posix_epoch then reads that local time as UTC, so in America/Denver every heartbeat lands 6 hours in the past and the client gives up on its first poll while the task completes on the server.

The fix stores naive UTC, which matches the column type and what SQLite and pg8000 already store.

• config_db_model.py adds _utc_now() (aware UTC now, tzinfo dropped) and uses it for the six task timestamp assignments.
• api/tasks.py makes _posix_epoch_to_db_timestamp return naive UTC too, so the getTasks time filters compare like with like.

There's no schema change and no migration. Rows written before the upgrade keep their old values, and they belong to tasks that already terminated. Nothing reads them to decide a timeout. CI's PostgreSQL service runs in UTC, which is why the matrix never caught it.

New unit tests in web/server/tests/unit/test_background_task_timestamps.py check that every task transition stores a naive UTC value and that the epoch conversion round trips. On master with only the test file added they fail, and with the fix they pass.

$ cd web && REPO_ROOT=.. BUILD_DIR=../build pytest -v server/tests/unit/test_background_task_timestamps.py
master 29100d4:
  ...test_abandoned_task_stores_naive_utc FAILED
  ...test_epoch_db_timestamp_round_trip FAILED
  ...test_task_transitions_store_naive_utc FAILED
  E   AssertionError: datetime.timezone.utc is not None
  3 failed, 1 warning in 1.95s
this branch:
  3 passed, 1 warning in 0.18s

$ make -C web test_unit_server
64 passed, 1 warning in 2.53s
$ make -k pycodestyle pylint
exit 0, every pylint run rated 10.00/10

End to end I ran a server built with BUILD_UI_DIST=NO make package against PostgreSQL 16.13 with timezone = 'America/Denver' and CODECHECKER_DB_DRIVER=psycopg2, then stored a one file clangsa analysis.

$ CodeChecker store reports -j 1 -n e2e --url http://localhost:8765/Default
master 29100d4:
  TaskTimeoutError: Task '5a00f8d8...' is still 'RUNNING', but did not have
  any progress for '6:00:00.411283' (21600.411283 seconds)!
  store_exit=1
  background_tasks: last_seen_at 06:58:28 while UTC now is 12:58:28;
  the task reached 'completed' on the server right after
this branch:
  Storing the reports finished successfully.
  store_exit=0
  background_tasks: completed, last_seen_at 12:59:27, UTC now 12:59:33

$ CodeChecker cmd serverside-tasks --url http://localhost:8765 --enqueued-after 2026:10:7:12:59:0 --output json
one task, "enqueuedAt": "2026-10-07 12:59:27"
$ CodeChecker cmd serverside-tasks --url http://localhost:8765 --enqueued-after 2026:10:7:13:0:0 --output json
[]

The same store against a SQLite product on this branch also exits 0. I didn't run make -C web test_matrix_psql_psycopg2 or the functional suites locally. CI runs them, and its PostgreSQL runs in UTC, so the check above is the one that covers this bug.