detect regressions of pyright analysis time in CI

[jorenham]

ref: #775 (comment)

This could be done by comparing the outputs of e.g. uv run basedpyright --stats --verbose scipy-stubs of the base branch and the PR branch.
It would fail when the slowdown of a particular file would be greater than some constant threshold, e.g. 100 ms. Or perhaps instead of failing, it could post a mypy_primer-esque comment that reports the differences in performance, including any notable improvements.
Care should be taken to ensure that this will also work in case of removed or created files.