fix tailcall VM for gcc

[henderkes]

I recall seeing an issue about this, but it seems it just disabled tailcall vm selection instead of patching the only conflicting function

[henderkes]

cc @arnaud-lb for review

[arnaud-lb]

The related issue is #23054.

This builds successfully with your changes, with GCC 16, if we ignore a bunch of maybe-musttail-local-addr warnings. My memory is fuzzy, but this warning might have been an error in the past? We should verify that we are not breaking the build for older GCC versions. We could add a pragma to ignore that error as well.

Otherwise the change looks good. We should probably apply it to all compilers as this shouldn't hurt, and the code would be easier to follow.

Please target master only.

[henderkes]

GCC only added preserve_none support in gcc 16 so it can't break for earlier gcc versions, they couldn't arrive in the branch.

I'm not sure if we should be silencing the warnings, assuming they are legitimate it may be worth finding workarounds later on. But that's for another time.

[henderkes]

Otherwise the change looks good. We should probably apply it to all compilers as this shouldn't hurt, and the code would be easier to follow.

Technically it's an extra function call in the one-time runtime compilation on clang, but it really shouldn't matter for performance.

[arnaud-lb]

This looks good to me! Do you want to add NEWS entry? I will merge after that

[henderkes]

Done!