Feature: Implement domain specific treadmill and gaze physiological filters

[VaibhavGarg2003]

The experanto/filters/ directory currently has empty placeholder modules for treadmill_filters.py and gaze_filters.py. In systems neuroscience, it is standard practice to filter behavioral data based on the animal's physiological state before analysis.
For example:

Treadmill: Mouse visual cortex responses differ between running and resting states.
Researchers often want to analyze only one state at a time.

Gaze/Eye Tracking: Eye tracker data contains artifacts from blinks (pupil size drops to 0 or NaN) and saccades (rapid eye movements). These time windows should be excluded to ensure data quality.

I propose implementing the following filters, following the closure pattern established in common_filters.py:

treadmill_filters.py

• running_filter(threshold, smoothing_sigma): Returns valid TimeIntervals where the animal's treadmill speed exceeds a configurable threshold.
  OptionalGaussian smoothing prevents micro-fluctuations from splitting intervals.

gaze_filters.py

• blink_filter(vicinity, pupil_channel_idx): Detects blinks (pupil size drops to 0 or NaN) and excludes a configurable time window around each event.

Both filters will:

• Accept a SequenceInterpolator as input
• Return List[TimeInterval] of valid clean data
• Include comprehensive tests in a new tests/test_domain_filters.py
• Include Numpy style docstrings

[pollytur]

I would kindly ask to wait on this for now. We need to restructure what lives on dataset and experiment levels and this would affect the filtering logic.

[VaibhavGarg2003]

@pollytur ,This makes sense, I'll hold off on this pr until the restructuring is done. I'd love to contribute to the restructuring effort itself if possible. Is there an issue or draft pr I should follow to stay up to date on the new dataset/experiment architecture or something else?