fix: bug audit — broken test imports, training data overlap, cleanup

- Fix test_utils.py importing build_annotation_json_path from main
  instead of core.annotations (all 59 tests pass now)
- Fix get_training_data double-counting clips at same start_time
  in both positive and soft sets — subtract positive from soft
- Add cancel_flag to train_classifier so training can be interrupted
  between videos (TrainWorker passes self as cancel_flag)
- Remove orphaned core/export.py (was for deleted server API)
- Remove stale Dockerfile and docker-compose.yml (referenced server)
- Clean up leftover server/__pycache__ and client/ build artifacts
- Add torch to requirements.txt (was only mentioned in comments)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

core/db.py

@@ -283,6 +283,11 @@ class ProcessedDB:
             else:
                 soft_by_video.setdefault(fn, set()).add(st)
 
+        # Remove positive times from soft to avoid conflicting labels
+        for fn in pos_by_video:
+            if fn in soft_by_video:
+                soft_by_video[fn] -= pos_by_video[fn]
+
         result = []
         for fn in pos_by_video:
             sp = source_by_filename.get(fn, "")