Implement status handling for AI SAST false-positive detection flags
### Summary
Add support for setting the `status` of `vulnerability_flags` based on AI false-positive detection confidence.
New and updated flags will set:
* `detected_as_fp` when confidence ≥ 0.6
* `detected_as_not_fp` when confidence < 0.6
Manual dismissals continue to set `status = not_started`.
### Implementation
* Update `UpdateAiDetectionService` to derive `status` from `confidence_score`.
* Explicitly set `not_started` for manually dismissed flags.
* Add a batched background migration to update existing `vulnerability_flags` so their status matches the new logic.
### Result
AI detection flags will consistently reflect whether a vulnerability was detected as a false positive.
issue