fix(worker): scale slskd download backstop to album size (stop abandoning big albums)

The absolute download backstop was a flat _MAX_XFER_POLLS=1200 (~60min). A large
album (e.g. Drake "Scorpion", 521MB / 25 FLACs) from a slow-but-healthy serial
peer (~400KB/s, serving one file at a time) needs ~60-70min, so it was cut off
at 60min and — because an abandoned transfer is cancelled and re-enqueued from
scratch — re-downloaded from zero every attempt, never finishing and eventually
going needs_attention at the attempt cap.

Replace the flat cap with _max_polls(total_bytes): budget the backstop at a
conservative floor throughput (~64KB/s), clamped to [20min, 2h]. Scorpion now
gets the full 2h ceiling instead of 60min. The 90s stall timeout is unchanged,
so a truly dead/queued peer is still abandoned fast.

Follow-up (not in this change): resume across retries (skip files slskd already
completed) so an abandoned large download doesn't restart from zero.

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

worker/tests/test_slskd_cancel.py

@@ -8,8 +8,12 @@ import pytest
 import lyra_worker.adapters._slskd as slskd_mod
 from lyra_worker.adapters._slskd import (
     _ETA_CAP_SECONDS,
+    _MAX_XFER_POLLS,
+    _MIN_XFER_POLLS,
+    _POLL_SECONDS,
     SlskdClient,
     _eta_seconds,
+    _max_polls,
     _parse_search_responses,
 )
 
@@ -162,6 +166,25 @@ def test_eta_seconds_unknown_cases():
     assert _eta_seconds(1000, 1000, 100.0) is None    # already complete
 
 
+def test_max_polls_scales_backstop_to_album_size():
+    # A 521MB album (Drake "Scorpion", 25 FLACs) from a slow serial peer must get far more than
+    # the old flat 1200-poll/60min budget, or it gets abandoned mid-download and restarts from zero.
+    scorpion_bytes = 521_000_000
+    polls = _max_polls(scorpion_bytes)
+    assert polls > 1200                              # more headroom than the old flat cap
+    assert polls * _POLL_SECONDS >= 90 * 60          # at least ~90 min of wall-clock budget
+
+    # A tiny single stays on the floor, not zero.
+    assert _max_polls(3_000_000) == _MIN_XFER_POLLS
+    assert _max_polls(0) == _MIN_XFER_POLLS
+
+    # An enormous set is clamped to the ceiling (bounds worst-case slot hold).
+    assert _max_polls(50_000_000_000) == _MAX_XFER_POLLS
+
+    # Monotonic: bigger album never gets a smaller budget.
+    assert _max_polls(200_000_000) <= _max_polls(400_000_000)
+
+
 def test_eta_seconds_capped_when_peer_nearly_stalls():
     # A near-stalled peer (tiny speed) would yield an ETA far past int4, overflowing the
     # downloadEtaSeconds column write. The estimate must be clamped to a safe ceiling.