Correct canonical psychophysical kernel timing

[rojasgabriel]

Summary

• put Bpod stimulus and observation times in the same trial-relative coordinate system
• require NIDAQ flashes and port events to align to their Bpod predictions within 100 ms, excluding incomplete trials instead of mixing clocks
• reduce pooled kernels to the fixed-width Odoemene Equation 5 model with opaque integer configuration IDs for center-exit and response fits
• replace the disposable legacy kernel tables through an explicit migration and strengthen timing, schema, and migration tests

Test plan

• uvx ruff check .
• uvx ruff format --check .
• uv run python -m unittest discover -s tests -v
• compare aligned Bpod and NIDAQ extraction on a live session
• reset and repopulate the live disposable kernel tables after merge

Made with Cursor