HIR: Move lifetime elision to HIR lowering

[nrc]

If we can

[arielb1]

@nrc

it is somewhat hard, as elision can occur in type aliases. This of course all depends on moving resolution.

[eddyb]

This seems to assume lifetime elision is a syntactical sugar, when it's actually semantic sugar (it depends on lifetimes in definitions of types).

[nrc]

I believe at least some elision is purely syntactic - that could be moved here and the type-dependent stuff left alone. Not sure about the aliases issue.

[jonas-schievink]

This of course all depends on moving resolution.

As far as I can tell, this was done in #33443. Can this now be fully implemented?

[eddyb]

@jonas-schievink Hmm, I suppose you can always look at lifetime parameter definitions in custom user types, so this would work. However, the current implementation takes advantage of the Debruijn index representation of late-bound lifetimes to avoid actually touching the lifetime parameter definitions.

[eddyb]

#39305 did as much as I thought was reasonable, anything more would really pollute lowering.
@nrc are you happy with that progress, for closing this issue?