Scope MLflow experiment names to include user ID (Task 14.2)

- Updated FastAPI /training/start endpoint to extract X-User-ID header via get_user_id() dependency
- Modified _run_training_background to accept and use user_id parameter
- Added MLflow experiment setup with user scoping: experiments are named user_{user_id}_training when user_id is provided, falling back to default experiment name otherwise
- Updated database record insertion to store scoped experiment name
- Updated training/train.py train() function to accept user_id parameter and use it for experiment naming
- Mark task 14.2 as complete in tasks.md

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

openspec/changes/user-accounts/tasks.md

@@ -82,7 +82,7 @@
 ## 14. ML Service User Scoping
 
 - [x] 14.1 `[haiku]` Update FastAPI service to read `X-User-ID` header from incoming requests
-- [ ] 14.2 `[haiku]` Scope MLflow experiment names to include user ID (e.g., `user_{uuid}_training`)
+- [x] 14.2 `[haiku]` Scope MLflow experiment names to include user ID (e.g., `user_{uuid}_training`)
 - [ ] 14.3 `[sonnet]` Scope training run queries in FastAPI to filter by user ID
 
 ## 15. Documentation & Deployment