Add note about limiting thread oversubscription by default

[jrbourbeau]

I suspect users interested in this topic would also find it useful to know these environment variables are set by default when using the distributed scheduler. This is a small follow-up to dask/distributed#7177

[jrbourbeau]

Thanks for reviewing @ncclementi!

Noting that test failures are unrelated -- see #9597

[jrbourbeau]

Thanks for reviewing @ncclementi!

Noting that test failures are unrelated -- see #9597

[jakirkham]

Wonder if we should be using threadpoolctl, given they help users do this sort of thing

[jrbourbeau]

Ah, neat. I've been meaning to read up more on threadpoolctl -- what does it get us beyond what setting the *_NUM_THREADS environment variables do?

[jakirkham]

It's also able to inspect how BLAS & LAPACK are configured and share that with the user. The package is pretty barebones, but maybe that makes it easier to adopt

Suppose the other option would be looking at using loky for our executors, but that might be a bigger change than we want to make (though it has some other perks)