add a tvb_kernels package

[maedoc]

This is a WIP to add a computation kernels package, focused on moving known hotspots from TVB's Python code into C. The set of kernels is for the moment purposefully small, covering just the delay-coupling calculation, in order to move through the whole integration, packaging & deploy process prior to achieving broader coverage.

While the kernels package will have wheels on every relevant platform, a basic NumPy fallback implementation will be maintained, with a primary goal of testing correctness.

The PR add a workflow to do standard pip-based build as well as a spack based build in anticipation of ebrains integration, but here's a few things required before a merge:

• [ ] add use of coupling kernels in TVB w/ benchmarks
• [ ] create a pypi project to submit wheels to
• [ ] add ci build wheels for all platforms
• [ ] add a Spack package definition
• [ ] add dep on this package to tvb library (both pip and spack) & test
• [ ] test within Ebrains & TVB Dockerfile builds
• [x] clean up commit history

[maedoc]

actions are being run over on my fork https://github.com/maedoc/tvb-root/actions tho I could push the branch here.