feat(worker): chunk the library scan + reconnect on DB loss

Two worker-loop robustness fixes:

- Scan chunking: maybe_run_scan advances the library scan by at most one
  scan_chunk (default scan.chunkSize=25 albums) per loop iteration, so a
  large library no longer blocks job-claim/monitor for minutes-to-hours.
  Progress persists in scan.inProgress/scan.cursor/scan.progress and is
  resumable; scan.result (unchanged format) is written on completion.

- DB reconnect: wrap the loop body in `except psycopg.OperationalError`
  (AdminShutdown subclasses it) to close the dead handle and reconnect via
  wait_for_db() in-process, instead of crashing and relying on the
  container restart policy. Verified via a backend-termination test.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

worker/tests/test_db_reconnect.py

@@ -0,0 +1,35 @@
+import psycopg
+import pytest
+
+from lyra_worker.db import connect, wait_for_db
+
+
+def test_terminated_backend_raises_operational_error_then_reconnects(conn):
+    """The run_forever reconnect boundary catches psycopg.OperationalError and re-runs
+    wait_for_db(). This exercises both primitives: a server-side backend termination (the
+    real AdminShutdown case) surfaces as OperationalError, and wait_for_db reconnects."""
+    victim = connect()
+    try:
+        with victim.cursor() as cur:
+            cur.execute("SELECT pg_backend_pid()")
+            pid = cur.fetchone()[0]
+        with conn.cursor() as cur:  # terminate the victim's backend, as a DB restart would
+            cur.execute("SELECT pg_terminate_backend(%s)", (pid,))
+        conn.commit()
+        with pytest.raises(psycopg.OperationalError):  # the exact type the loop catches
+            with victim.cursor() as cur:
+                cur.execute("SELECT 1")
+                cur.fetchone()
+    finally:
+        try:
+            victim.close()
+        except Exception:
+            pass
+
+    fresh = wait_for_db()  # reconnect exactly as run_forever does
+    try:
+        with fresh.cursor() as cur:
+            cur.execute("SELECT 1")
+            assert cur.fetchone()[0] == 1
+    finally:
+        fresh.close()

worker/tests/test_scan_trigger.py

@@ -1,5 +1,5 @@
 from lyra_worker.adapters.fakes import FakeMbBrowser
-from lyra_worker.main import _run_scan
+from lyra_worker.main import maybe_run_scan
 from lyra_worker.types import MBTarget
 
 
@@ -11,7 +11,7 @@ class FakeProbe:
 
 class FakeResolver:
     def resolve(self, artist, album):
-        return MBTarget(artist=artist, album=album, rg_mbid="rg1", artist_mbid="a1")
+        return MBTarget(artist=artist, album=album, rg_mbid=f"rg-{album}", artist_mbid="a1")
 
 
 def _set(conn, key, value):
@@ -31,15 +31,55 @@ def _get(conn, key):
         return row[0] if row else None
 
 
-def test_run_scan_writes_result_and_clears_flag(conn, tmp_path):
-    (tmp_path / "John Mayer" / "Continuum (2006)").mkdir(parents=True)
-    (tmp_path / "John Mayer" / "Continuum (2006)" / "01.flac").write_bytes(b"")
+def _album(tmp_path, artist, folder):
+    d = tmp_path / artist / folder
+    d.mkdir(parents=True)
+    (d / "01.flac").write_bytes(b"")
+
+
+def _cfg(conn):
+    from lyra_worker.config import get_config
+    return get_config(conn)
+
+
+def test_scan_completes_in_one_chunk_writes_result_and_clears_flag(conn, tmp_path):
+    _album(tmp_path, "John Mayer", "Continuum (2006)")
     _set(conn, "scan.requested", "true")
 
-    _run_scan(conn, FakeResolver(), FakeProbe(), FakeMbBrowser(), dest_root=str(tmp_path))
+    # one album, default chunk size 25 -> starts + finishes in a single call
+    maybe_run_scan(conn, FakeResolver(), FakeProbe(), FakeMbBrowser(), _cfg(conn), dest_root=str(tmp_path))
 
-    assert _get(conn, "scan.requested") == "false"      # flag cleared
+    assert _get(conn, "scan.requested") == "false"          # flag cleared
+    assert _get(conn, "scan.inProgress") == "false"         # scan finished
     assert "imported" in (_get(conn, "scan.result") or "")  # summary written
     with conn.cursor() as cur:
         cur.execute('SELECT count(*) FROM "LibraryItem"')
         assert cur.fetchone()[0] == 1
+
+
+def test_scan_chunks_across_iterations_and_totals_correctly(conn, tmp_path):
+    for i in range(3):
+        _album(tmp_path, "Artist", f"Album {i}")
+    _set(conn, "scan.requested", "true")
+    _set(conn, "scan.chunkSize", "2")
+
+    # first call starts the scan and processes chunk 1 (2 albums), leaving it in progress
+    maybe_run_scan(conn, FakeResolver(), FakeProbe(), FakeMbBrowser(), _cfg(conn), dest_root=str(tmp_path))
+    assert _get(conn, "scan.inProgress") == "true"
+    assert _get(conn, "scan.requested") == "false"
+    assert _get(conn, "scan.progress") == "2/0"
+
+    # second call resumes from the cursor, finishes the last album
+    maybe_run_scan(conn, FakeResolver(), FakeProbe(), FakeMbBrowser(), _cfg(conn), dest_root=str(tmp_path))
+    assert _get(conn, "scan.inProgress") == "false"
+    assert _get(conn, "scan.result") == "imported 3, skipped 0"
+    with conn.cursor() as cur:
+        cur.execute('SELECT count(*) FROM "LibraryItem"')
+        assert cur.fetchone()[0] == 3
+
+
+def test_maybe_run_scan_idle_when_not_requested_or_in_progress(conn, tmp_path):
+    # no flags set -> no-op, no result written
+    maybe_run_scan(conn, FakeResolver(), FakeProbe(), FakeMbBrowser(), _cfg(conn), dest_root=str(tmp_path))
+    assert _get(conn, "scan.result") is None
+    assert _get(conn, "scan.inProgress") is None