feat: added from_idtracker to generate bbox annotations from idtracker.ai (#8)

[GauravSRC]

Closes #8

Summary

Adds ethology/io/annotations/load_idtracker.py with a public function
from_idtracker that converts idtracker.ai output files into a valid
ethology bounding box annotations xr.Dataset that can be reviewed
by a human and then exported with the existing save_bboxes.to_COCO_file.

New files

File	Purpose
ethology/io/annotations/load_idtracker.py	from_idtracker + two private helpers
tests\test_unit\test_io_annotations\test_load_idtracker.py	40 unit tests, synthetic data only

How it works

Inputs

Parameter	Type	Notes
trajectories_path	.npy path	shape (n_frames, n_animals, 2)
frame_indices	list[int]	0-based, duplicates removed silently
bbox_size	(width, height)	required when blobs_collection_path is None
blobs_collection_path	.pkl path	optional; uses actual blob bboxes when present
images_dir	path	stored in attrs, not required for computation

Output -- xr.Dataset validated by ValidBboxAnnotationsDataset:

• position (image_id, space, id)-- centroid (x, y) in pixels
• shape (image_id, space, id) -- (width, height) in pixels
• category (image_id, id) -- 1-based animal identity

Frames where an animal is undetected (NaN centroid in trajectories / no
blob with matching identity) produce NaN in position/shape and
-1 in category, consistent with the NaN-padding convention used
throughout the rest of the codebase.

Testing

All tests use synthetic numpy arrays and unittest.mock.MagicMock
blobs -- no external test-data files or network access are needed.

Checklist

• New module follows the exact same import / decorator pattern as load_bboxes.py
• Output is validated by @_check_output(ValidBboxAnnotationsDataset)
• Full docstring with Parameters, Returns, Raises, Notes, Examples
• All private helpers have their own docstrings and unit tests
• No changes to existing files