Update to wasi-sdk-30

[alexcrichton]

Seeing if any issues arise...

[alexcrichton]

After a-bit-too-much-debugging I've learned:

• The failure here is due to a Python-semantics issue where an object is alive when it shouldn't be.
• That's because of this logic where deallocs are deferred when Python thinks it's out-of-stack
• Python thinks it's out-of-stack due to two reasons:
  • The root issue is this commit, a patch for this fork of Python. I don't think that's necessary any more and deleting that commit makes everything pass.
  • Between wasi-sdk-27 and wasi-sdk-30 Fix a few users of single-thread TLS WebAssembly/wasi-libc#618 landed. This enabled a function which python uses to probe the stack size of the current thread meaning that the stack size calculations differed from before (got more accurate I believe).

By reverting dicej/cpython@5522080 I got tests to pass locally with wasi-sdk-30. @dicej can you make a new branch in your fork with that commit missing?

[dicej]

Thanks for figuring this out, @alexcrichton! I'll update my CPython fork and then update the submodule in this PR.