Remove __syscall_pause

[sbc100]

We don't support this syscall and musl already has a fall back based on poll/ppoll.

Note that the return value for pause() now changes to zero (from ENOSYS), because it is now based in poll internally and currently our __syscall_poll implementation returns zero when it would otherwise need to block. I have another change out the potentially change that (#27049) but we can consider that separately to this change.

See #27036 and #27038

[sbc100]

I simplified this PR again so that pause now returns 0.

I explained why I think that is OK in the PR description and we can followup with #27049 if we decide we prefer these functions to fail rather than lie about blocking.

[kripken]

I'm not opposed to this change, but let's mention it in the changelog?

[sbc100]

What do you think about updating poll in #27049 to return non-zero when it needs to block but cannot.

If we land #27049 it would have the side effect of restoring this pause() behaviour?

[sbc100]

Let me phrase this as choice of two options that I'd like us to choose between for the next release:

1. pause() syscall goes from EINTR to silent success. poll() continues to return silent success when it should really block.
2. Change poll() to that it no longer silently return success and instead returns EINTR when it cannot block but needs to. If we do this then the behaviour of pause() will continue to be EINTR (since pause if uses poll to block).

[kripken]

I feel I don't have enough of a sense of how user code uses this syscall to pick between them. So I don't feel strongly.

But if we change a syscall or libc function, I think it's worth mentioning that breaking change.

[sbc100]

I'm thinking of going with #27049 which will effecitivly revert the user-visible change in this PR, but I will update the ChangeLog not and then remove if I land #27049