CI is slow

[ludfjig]

We should investigate what it takes to speed up CI. I have some suggestions, but I'm sure more can be done. Feel free to add anything I forgot

• Benchmarks on windows runner is very slow compared to linux. We should consider decreasing number of samples needed on windows
• The same wit tests run twice, once for GUEST=c, and once for GUEST=rust, but in this case the tests are identical, so no need to rerun.
• We should be able to trim down some dev-dependencies to improve compile times using cargo features
• -Can we cache rust artifacts in github actions cache to improve compile times?

[simongdavies]

One quick win here is there is one windows benchmark that takes a long time , fixing just this will trim minutes off CI:

Benchmarking guest_functions/guest_call_with_large_parameters: Collecting 100 samples in estimated 1193.1 s (100 iterations)

[jsturtevant]

#632 was a short term fix for speeding up windows

[syntactically]

I find this particularly irritating when using the rebase workflow, which on GH with our configuration requires manually rebasing and erases PR approvals when main changes under you. This means that when trying to merge anything you need to keep a close eye on CI so that you can merge it as soon as CI finishes before something else updates main under you and starts the whole thing over again. Either making CI faster, or having a way to auto-approve rebases which do not actually change the diff, would do a lot to reduce the amount of time I spend staring at CI waiting for it to finish so that I can push to main before someone else does.