request: report the number of compilation errors

[hinaria]

[Mark-Simulacrum]

Nominating for compiler team: we should make a decision here and stick to it, and I'm inclined to say that restoring the old functionality also leads to reports of incorrect numbering. If we can fix the numbering (easily) then I'm all for it though, but last I looked this wasn't an easy task.

[Ixrec]

If completely fixing the numbers is non-trivial, is it feasible/acceptable to report "There were at least 10 errors"? Or is the count so buggy we don't even know a minimum?

[Mark-Simulacrum]

IIRC, I saw an issue recently about the returned value being 3 where it should've been 1, so I don't think that's true. It's possible that was due to old code though.

[nikomatsakis]

triage: P-medium

My feeling is that if we can get a correct count, that would be good. This doesn't seem too hard in principle, but I guess there are... obstacles.