Make the worker batch timeout configurable via LIDAR_BATCH_TIMEOUT

The pool of tile workers used to cancel every remaining tile after a
hardcoded 2-hour wall clock, silently truncating large batches (a
670-tile completion run lost its last 348 tiles that way). The timeout
now defaults to unlimited and can be capped per deployment with the
LIDAR_BATCH_TIMEOUT environment variable (seconds); the local worker
compose sets it to 6 hours.

💘 Generated with Crush

Assisted-by: Crush:glm-5.2

lidar_pipeline/tests/test_pipeline.py

@@ -440,3 +440,21 @@ class TestResolveWorkers:
         for _ in range(4):              # 4th call: empty queue
             pipeline._init_worker_slot(q)
         assert calls == [("gpu", 1), ("cpu",)]
+
+
+class TestBatchTimeout:
+    def test_env_var_parsing(self, monkeypatch):
+        """LIDAR_BATCH_TIMEOUT: unset/0/invalid = unlimited, N seconds = N."""
+        from lidar_pipeline.pipeline import _batch_timeout_s
+        monkeypatch.delenv("LIDAR_BATCH_TIMEOUT", raising=False)
+        assert _batch_timeout_s() == 0.0
+        monkeypatch.setenv("LIDAR_BATCH_TIMEOUT", "")
+        assert _batch_timeout_s() == 0.0
+        monkeypatch.setenv("LIDAR_BATCH_TIMEOUT", "0")
+        assert _batch_timeout_s() == 0.0
+        monkeypatch.setenv("LIDAR_BATCH_TIMEOUT", "3600")
+        assert _batch_timeout_s() == 3600.0
+        monkeypatch.setenv("LIDAR_BATCH_TIMEOUT", "-5")
+        assert _batch_timeout_s() == 0.0
+        monkeypatch.setenv("LIDAR_BATCH_TIMEOUT", "abc")
+        assert _batch_timeout_s() == 0.0