Add comprehensive thread-worker unit coverage; harden CSVImportThread ROI field/NaN handling

[Copilot]

This PR adds focused unit coverage for previously untested thread workers under matchypatchy/threads, with priority on import_thread.py and animl_thread.py. While building the tests, it also addresses two CSV import edge cases surfaced by coverage (ROI field access mode mismatch and favorite/individual null handling).

• Import thread test suite (tests/test_threads/test_import_thread.py)
  • Adds class-based tests for:
    • CSVMigrateThread: required-column validation, row processing, ref-dict caching, missing-file skip/error collection, interruption/finish behavior.
    • CSVImportThread: common base-dir detection, filepath grouping behavior, old (bbox1-4) and new (bbox_x/y/w/h) bbox formats, value conversion with None/NaN, ROI/media thumbnail persistence, interruption/finish behavior.
    • FolderImportThread: station/camera extraction by path index, default station creation (station_level=0), interruption and finish semantics.
    • BasePathUpdateThread: update_base_dir invocation semantics and progress/finished signal behavior.
• ANIML thread test suite (tests/test_threads/test_animl_thread.py)
  • Adds class-based tests for:
    • BuildManifestThread: manifest build call contract and manifest signal payload.
    • VerifyNewBaseDirsThread: hash-diff logic across update tuples and not_in_db / not_in_new_directory / finished signaling.
    • AnimlThread: initialization from DB state, image/video detection flows, ROI bbox updates for unresolved ROIs, thumbnail saves, progress/prompt signaling, detector=None behavior, interruption handling.
• Test runtime support
  • Extends tests/conftest.py with minimal QThread/pyqtSignal stubs to reliably exercise worker run() paths in headless unit tests.
• Targeted production hardening in CSVImportThread
  • Adds _roi_value() to support ROI extraction when selected columns are passed as either names or positional indices.
  • Normalizes null semantics:
    • favorite now defaults to 0 when conversion yields None.
    • individual() now returns None for missing/NaN names instead of attempting DB lookup/insert.

def _roi_value(self, roi, column):
    if column == "None" or column is None:
        return None
    if isinstance(column, int):
        return roi[column]
    return getattr(roi, column, None)

def favorite(self, roi):
    if self.selected_columns["favorite"] != "None":
        favorite = self._convert_to_int(self._roi_value(roi, self.selected_columns["favorite"]))
        return favorite if favorite is not None else 0
    return 0