test: Re-enable mips(el)-linux(-musl) tests.

[alexrp]

Andrew mentioned that LLVM's MIPS backend has O(N^2) codegen. Just want to take this for a spin and see if that remains the case.

[linusg]

Unfortunately that's still the case, at least I observe excessive build times when building Kiesel for mips-linux-musl (~100-120s vs ~30s for mips64-linux-musl and x86_64-linux-musl).

There's #18872 (and dddddcf) for RISC-V, we might want to have a separate issue for MIPS depending on the outcome of this PR.

[alexrp]

Hmmmmmm... there's enough fluctuation in run times across PRs that it's a bit hard to tell how bad this is. Some of my other PRs have run times that look almost identical to this, but then others are anywhere between 20-40 minutes faster. Presumably it completely depends on how busy CI is at any given time, and it seems to be quite busy right now. That said, based on my local testing, I think it's fair to assume that this will at least add 20 minutes to run times.

Totally understand if this is not acceptable; if so, I can just separate out the two commits that disable failing tests and PR those independently.

[andrewrk]

It's most noticeable when you test locally with zig build test-std.

But, OK, let's try it for now. Maybe an option to skip those can be added which can be useful for local testing.

[alexrp]

That seems like a good idea. I would even say it'd be fine for these slower targets to be off by default, with CI explicitly opting into testing them.