Add compiling linux as a benchmark

[derSteFfi]

This adds compiling linux as a benchmark.
I opted for the latest stable version as of right now. It's building a 'allnoconfig' to not run too long. It's building in RAM to reduce file I/O overhead. I have no idea how to realize that on platforms other than *nix, though.

Comments and feedback welcome :)

[mjp41]

Thanks for contribution to mimalloc-bench. I am a little worried about compiling Linux completely in memory for the CI machines. Last time I checked they only had 7GiB of RAM. I have set it running, and lets see what happens.

[jvoisin]

This is working, amazing.

I think we should investigate why Alpine is taking more than 2h before merging this.

[mjp41]

I am not sure this is actually running in CI. The set tests_all3 which linux is added to does not seem to be part of the CI job.

[derSteFfi]

It's not. I wasn't thinking about the CI, I just didn't want to force it into allt, given it runs for about 2 minutes. Shall I move it to test2?

[mjp41]

It's not. I wasn't thinking about the CI, I just didn't want to force it into allt, given it runs for about 2 minutes. Shall I move it to test2?

Not sure, this is probably a deeper question about what CI is checking. At the moment, it doesn't check all the benchmarks which is a bit of a problem, but it also takes a long time already.

I don't think having a two minute task in allt is ideal, so I think it is in the right place. But it would be good to test more benchmarks in CI.

I'll think a bit about how we might speed up CI. I am wondering if we should containerize some stuff, so we can reuse artifacts, and then run the benchmarks across more CI machines.