Rollup merge of #147729 - lqd:polonius-diagnostics, r=jackh726

ignore boring locals when explaining why a borrow contains a point due to drop of a live local under polonius

Polonius liveness has to contain boring locals, and we ignore them in diagnostics, to match NLL diagnostics that never involve any boring locals. When explaining why a borrow contains a point, I ignored these boring locals when it was due to a use of a live var, but forgot to do so when the cause was because of a drop of a live var.

This is what was causing the last two (known) diagnostics differences under the polonius compare-mode:
- `tests/ui/dropck/dropck_trait_cycle_checked.rs`
- `tests/ui/unboxed-closures/unboxed-closures-failed-recursive-fn-1.rs`

r? `@jackh726`

Author: matthiaskrgr