feat: curio ffiselect: Isolate gpu calls in a subprocess

[snadrus]

Related Issues

The GPU picker library "ffiselect" will run ffi calls on an assigned GPU using a subprocess. It does not respect deadlines, but will survive fatal behavior of the subprocess.

This PR also removes localWorker (tech debt) from curio and copies code from there and the sealer into Curiosrc near the usage.

Proposed Changes

Additional Info

Checklist

Before you mark the PR ready for review, please make sure that:

• Commits have a clear commit message.
• PR title is in the form of of <PR type>: <area>: <change being made>
  • example: fix: mempool: Introduce a cache for valid signatures
  • PR type: fix, feat, build, chore, ci, docs, perf, refactor, revert, style, test
  • area, e.g. api, chain, state, market, mempool, multisig, networking, paych, proving, sealing, wallet, deps
• If the PR affects users (e.g., new feature, bug fix, system requirements change), update the CHANGELOG.md and add details to the UNRELEASED section.
• New features have usage guidelines and / or documentation updates in
  • Lotus Documentation
  • Discussion Tutorials
• Tests exist for new functionality or change in behavior
• CI is green

[snadrus]

@magik6k please review my approach.

1. how do I avoid calling WindowPoSt in a loop?
2. WInningPoSt:

• Should I use GenerateSingleVanillaProof from the FFI Select wrapper?
• Is there a better winningPoSt low-level call?

3. Parallelism: where should I have it? I removed it from WindowPoSt.

[magik6k]

I feel quite strongly that we're passing those calls one layer of abstraction above where we should be passing them.

[github-actions]

All checks have completed

❌ Failed Test / Test (itest-curio) (pull_request)
❌ Failed Test / Test (itest-net) (pull_request)

[magik6k]

The linter does appear to be correctly complainy about how we return errors from the ffiselect wrapper.

Some minor things to address, overall looks really good, can't wait to give this a try!

[snadrus]

Your changes look good @magik6k. I also added an always-ran "test" to verify everything looks sharp. I realize that part of that could be a unit test instead, but I'm not sure if that's working well at the moment.

[magik6k]

Looks good now, and works on my machine. Flaky curio test is not from this PR, will fix in a followup