Add First Pass Tracking Functionality

[emdavis02]

PR Description

The sleap-mot repo currently supports tracking pose-estimated videos either:

1. From scratch (no prior identity information), or
2. With initial tracklets produced from first-pass tracking.

First-pass tracking uses visible features or external signals to assign partial identities (tracklets) to animals when those features are available. For example:

• Visual features like fur color can distinguish animals when they're unobstructed.
• RFID pings can be matched to individuals when a signal is detected.

Previously, these feature-based tracking methods were implemented in scattered Jupyter notebooks. The goal of this PR is to incorporate them directly into the sleap-mot repository, enabling it to function as a complete identity tracking pipeline, from raw pose data to globally consistent tracks.

---

Key Changes

• Introduced a general FeatureTracker base class
  This abstract base class defines shared functionality and interfaces for all feature-based tracking approaches. It includes:
  • Utility functions for tracklet management
  • A common structure for integrating with the broader tracking pipeline
  • An abstract method track() that must be implemented differently in each subclass depending on the feature type

Implemented subclasses:

• RFIDFeatureTracker
  This subclass demonstrates how specific feature-based methods (in this case, RFID-based) can be cleanly integrated using the inherited structure and shared utilities from FeatureTracker.

• FurColorFeatureTracker

• Test functions will be written to test the functionality of these new classes in tandem with the tracker class (second step tracking)

---

Outcome

This PR formalizes and integrates feature-based first-pass tracking into the core sleap-mot package. It lays the groundwork for a modular, extensible identity tracking system, where multiple feature types (e.g., color, RFID, embeddings) can be easily added and composed.

[codecov]

Codecov Report

❌ Patch coverage is 2.48500% with 1138 lines in your changes missing coverage. Please review.
✅ Project coverage is 20.81%. Comparing base (2cb2db0) to head (4dc8d41).
⚠️ Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
sleap_mot/feature_tracker.py	0.00%	994 Missing ⚠️
sleap_mot/tracker.py	26.19%	62 Missing ⚠️
sleap_mot/utils.py	12.50%	49 Missing ⚠️
sleap_mot/eval.py	0.00%	33 Missing ⚠️

❗ There is a different number of reports uploaded between BASE (2cb2db0) and HEAD (4dc8d41). Click for more details.

HEAD has 2 uploads less than BASE

Flag	BASE (2cb2db0)	HEAD (4dc8d41)
	5	3

Additional details and impacted files

@@             Coverage Diff             @@
##             main      #33       +/-   ##
===========================================
- Coverage   44.73%   20.81%   -23.92%     
===========================================
  Files           8        9        +1     
  Lines         930     2066     +1136     
===========================================
+ Hits          416      430       +14     
- Misses        514     1636     +1122     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[talmo]

Could do something like:

cols = [*[f"{node_name}.x", f"{node_name}.y", f"{node_name}.score"] for node_name in skeleton.node_names]

pd.DataFrame(inst.numpy(scores=True).reshape(-1), columns=cols)