I've run into frustrating stack overflows in seemingly trivial non-recursive code, so I appreciate this effort!

I wonder if it might be simpler to track stack sizes statically instead of using runtime instrumentation. What I mean is, for example on x86_64, the function prologue has a stack reservation in the form of `sub rbp, 0x168`. So we can easily tell that the function uses 0x168 bytes of stack space. Just add those numbers up for every function in a crate, and you have a score. Track that number over time for a set of common crates.