[compiler-rt] Restore unsigned value in struct, use enum only in function

[mikolaj-pirog]

Typed enums are c23 features and are too new to be used. This PR restores the types in the __processor_model struct back to unsigned int, removes typed enums, and uses the enum in the function as a variable that's later assigned to a global in order to prevent errors fixed initially here: #164713

See #165034 for more background

[compnerd]

Likewise

[mikolaj-pirog]

I don't follow, I think one of your comments wasn't posted

[compnerd]

Ugh, silly GitHub. I was thinking early return might be useful here.

[e-kud]

I won't block the PR but I don't like this kind of refactor. Previously it was a pure static function that touched no context. Now it implicitly modifies the global variable. References to __cpu_type and __cpu_subtype were additional return values.

[mikolaj-pirog]

I understand the downsides, but I believe in this file it's not gonna be a problem, especially how the function is local.

I can refactor later to not use the global like that, but the previous solution of passing a pointer to some fields of the global isn't much different than what the code is doing now

[yuxuanchen1997]

I commented on my original PR that passing pointers to field while assuming a different signedness is not good. What about passing pointers to entire struct? That would required the struct to be named but I think we should be fine by that?

[compnerd]

Yeah, passing the pointer to the struct would be my preference as well.