stage2: comptime function calls and inline function calls

[andrewrk]

See commit messages for more details.

Follow-up issues to tackle:

• Adding compile errors to the callsite instead of the callee Decl.
  • add test for this
• Proper error notes for "called from here". "called from here" error hints when a compile error occurs in a function being inlined or comptime called #7668
• Branch quotas for both comptime fn calls and inline function calls.
• Complete out the debug TZIR dumping code. Complete out the debug AIR dumping code #7670
• Don't redundantly generate ZIR for each inline/comptime function call. Instead we should add a new state enum tag to Fn.
• add more test cases
• caching of comptime function calls (required for generic types to work) caching of comptime function calls #7669

[g-w1]

Should you change this to make it if (b.is_comptime or inst.kw_args.modifier == .compile_time) ?
Or will this come later with the hashing and determining if something can be called at comptime?

[andrewrk]

Yep, good call. I'm actually looking into right now what it would look like to support inline function calls too, because I think the problem is related.

[g-w1]

Also, to note, this segfaults the compiler with no stacktrace (pretty sure stack overflow):

export fn _start() noreturn {
    const v = x;
    unreachable;
}

const x = rec(3);

fn rec(x: u32) u32 {
    return rec(x);
}

Recursion support will probably come with the branch eval quota, but it was surprising to me that there was no trace.

Another compelling reason to cache the astgen, besides having something run at comptime and runtime is comptime recursion. If im not mistaken, every time this function is called at comptime, astgen is redundantly run on it.

[andrewrk]

stack trace on stack overflow is #1616

yep, I have a bunch of unpushed code that addresses the ZIR caching, almost ready to push!