feat(testing): add max_frames parameter to E2EHarness

Caps the iteration length (and the matching GT slice) when set, so CI
tiers can stay fast on multi-thousand-frame sequences like EuRoC MH_01
(3682 frames ≈ 3+ hours at 3-5s/frame). Also useful for eyeballing a
new adapter's first N frames before committing to a full run.

Three new harness tests cover truncation, explicit None, and over-large
limits. No change to existing adapters or downstream tests.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

src/gps_denied/testing/harness.py

@@ -45,11 +45,24 @@ class HarnessResult:
 
 
 class E2EHarness:
-    """Drives FlightProcessor from a DatasetAdapter; collects results."""
+    """Drives FlightProcessor from a DatasetAdapter; collects results.
 
-    def __init__(self, adapter: DatasetAdapter, flight_id: str = "e2e-flight") -> None:
+    `max_frames` caps the iteration (and the matching GT slice). Useful for:
+      - CI tiers where a full multi-thousand-frame sequence is too slow.
+      - Debugging a new adapter — eyeball the first N frames before committing
+        to a multi-hour run.
+    `None` (default) means "consume the full dataset".
+    """
+
+    def __init__(
+        self,
+        adapter: DatasetAdapter,
+        flight_id: str = "e2e-flight",
+        max_frames: Optional[int] = None,
+    ) -> None:
         self._adapter = adapter
         self._flight_id = flight_id
+        self._max_frames = max_frames
         self._estimates: list[tuple[int, Optional[tuple[float, float, float]]]] = []
 
     async def run(self) -> HarnessResult:
@@ -57,6 +70,10 @@ class E2EHarness:
         frames = list(self._adapter.iter_frames())
         gt_poses = list(self._adapter.iter_ground_truth())
 
+        if self._max_frames is not None:
+            frames = frames[: self._max_frames]
+            gt_poses = gt_poses[: self._max_frames]
+
         for frame in frames:
             image = self._load_or_synth_image(frame.image_path)
             result = await processor.process_frame(

tests/e2e/test_harness_smoke.py

@@ -35,3 +35,32 @@ async def test_harness_captures_ground_truth_as_enu():
     east_disp = result.ground_truth[-1, 0] - result.ground_truth[0, 0]
     # Allow 5% tolerance for the lat/lon → ENU conversion approximation
     assert abs(east_disp - 4.0) < 0.5
+
+
+@pytest.mark.asyncio
+async def test_harness_max_frames_truncates_iteration():
+    # Adapter says 10 frames; harness with max_frames=3 should stop at 3.
+    adapter = SyntheticAdapter(num_frames=10, fps=5.0)
+    harness = E2EHarness(adapter, max_frames=3)
+    result = await harness.run()
+    assert result.num_frames_submitted == 3
+    # GT aligned to the same truncation so downstream metrics match lengths
+    assert result.ground_truth.shape[0] == 3
+
+
+@pytest.mark.asyncio
+async def test_harness_max_frames_none_runs_full():
+    # Explicit None = no limit (same as omitting the parameter).
+    adapter = SyntheticAdapter(num_frames=4, fps=5.0)
+    harness = E2EHarness(adapter, max_frames=None)
+    result = await harness.run()
+    assert result.num_frames_submitted == 4
+
+
+@pytest.mark.asyncio
+async def test_harness_max_frames_larger_than_dataset_is_harmless():
+    # Limit above dataset size should not over-extend.
+    adapter = SyntheticAdapter(num_frames=4, fps=5.0)
+    harness = E2EHarness(adapter, max_frames=100)
+    result = await harness.run()
+    assert result.num_frames_submitted == 4