Ray: Balance movie tracing with dynamic chunks

Schedule ray-pool advancement in dynamic chunks of 32 slots to distribute frame-grouped active rays across workers while preserving stable endpoint destinations.

Add 1/4/16-thread lens-map equivalence coverage and document schedule comparisons with complete commands, input data and original logs. The 393-frame 64x36 production-source benchmark improves from 62.13 to 38.69 seconds with byte-identical PNGs and lens maps. Release and Debug test suites pass.

tests/test_camera_cli.py

@@ -12,8 +12,8 @@ BUILD = Path(sys.argv[1] if len(sys.argv) > 1 else 'build/Release').resolve()
 ENV = dict(os.environ, OMP_NUM_THREADS='4')
 
 
-def run(binary, *args, ok=True):
-    result = subprocess.run([str(binary), *map(str, args)], env=ENV,
+def run(binary, *args, ok=True, env=ENV):
+    result = subprocess.run([str(binary), *map(str, args)], env=env,
                             capture_output=True, text=True)
     if (result.returncode == 0) != ok:
         raise AssertionError(f'{binary.name} {args}: {result.returncode}\n{result.stderr}')
@@ -144,13 +144,28 @@ with tempfile.TemporaryDirectory(prefix='gr-camera-cli-') as directory:
             # former still needs refinement. Finishing the empty batch used
             # to abort the whole movie in generation 1.
             mixed = Path(__file__).parent / 'fixtures/schwarzschild_mixed_refinement.csv'
+            parallel_map = tmp / 'mixed-parallel.grlens'
             result = run(binary, *common, '--height', 36, '--fov-deg', 60,
                          '--refine-max-level', 3, '--observer-track', mixed,
                          '--movie-track-samples', '--frames-dir', tmp,
-                         '--frames-prefix', 'mixed', '--verbose')
+                         '--frames-prefix', 'mixed', '--verbose',
+                         '--lens-map-output', parallel_map)
             assert 'Ray trace generation 1: frame 0 added' in result.stderr
             assert 'Ray trace generation 0: frame 1 added' not in result.stderr
             for frame in range(2):
                 image_payload(tmp / f'mixed_{frame:06d}.png',
                               dimensions=(64, 36), allow_black=True)
+            # Thread scheduling must preserve endpoints, frame/sample IDs and
+            # the resulting adaptive mesh across the entire slab sweep.
+            for threads in (1, 16):
+                comparison_map = tmp / f'mixed-{threads}-threads.grlens'
+                run(binary, *common, '--height', 36, '--fov-deg', 60,
+                    '--refine-max-level', 3, '--observer-track', mixed,
+                    '--movie-track-samples', '--frames-dir', tmp,
+                    '--frames-prefix', f'mixed-{threads}-threads',
+                    '--lens-map-output', comparison_map,
+                    env=dict(ENV, OMP_NUM_THREADS=str(threads)))
+                assert parallel_map.read_bytes() == comparison_map.read_bytes(), \
+                    f'movie lens map changed with {threads} threads'
+            print('schwarzschild: movie lens map identical with 1, 4 and 16 threads', flush=True)
         print(f'{backend}: CLI checks passed; single/movie PNG identical, map max error {max_error:.3g}', flush=True)