Give `resolve_const_param_in_non_trivial_anon_const` more contextual information

[compiler-errors]

Hm, it would be really nice if we could actually pass a more rich description of the position that's disallowing the const, probably via the AnonConstKind, so that we don't have to say "here" but instead something like "in array repeat expressions" or "in const generic args" something.

Right now the errors just say "here", but I think we could pass in enough information in the resolver to know what position a non-trivial const arg actually is, which would greatly help discovery of the issue.

Originally posted by @compiler-errors in #142157 (review)

[dvtkrlbs]

@rustbot claim

[dvtkrlbs]

From what I see AnonConstKind is resolved at the late stage in the resolve. I am not sure how to feed more data into a diagnostic error that happens at ident phase. Another question is AnonConstKind is not public and defined in the late.rs file do I need to just use display to turn it into string or move it somewhere else and use it directly in the error definition with IntoDiagArg?