The deallocation span of locals that are deallocated by a tail call `become` is nonsense

[RalfJung]

Test code:

#![feature(explicit_tail_calls)]
#![allow(incomplete_features)]

fn g(x: *const i32) {
    let _val = unsafe { *x }; //~ERROR: has been freed, so this pointer is dangling
}

fn f(_x: *const i32) {
    let local = 0;
    let ptr = &local as *const i32;
    become g(ptr)
}

fn main() {
    f(std::ptr::null());
}

The deallocation span used to be

help: ALLOC was deallocated here:
  --> tests/fail/tail_calls/dangling-local-var.rs:LL:CC
   |
LL |     f(std::ptr::null());
   |     ^^^^^^^^^^^^^^^^^^^

now it is

help: ALLOC was deallocated here:
  --> tests/fail/tail_calls/dangling-local-var.rs:LL:CC
   |
LL |     let _val = unsafe { *x };
   |         ^^^^

Both of these make no sense. I think the ideal span would be to point at the become.