Consider making shadowing of type/lifetime parameters a lint

[nrc]

see #459

We accepted this RFC as a hard error, but it should probably be a configurable lint.

[ftxqxd]

This would be significantly less useful as a normal lint, because lints run after type checking and so on, and most problems caused by lifetime/type parameter shadowing give errors before the lints can run. It would probably have to be a different kind of lint that runs before type checking to be useful, and as far as I know we don’t have such a category of lints at the moment.

[nrc]

@P1start yeah, we discussed this in the meeting and that is why it is accepted as a hard error. In order to make this a lint, we need to improve the lint infrastructure to allow lints to run before type checking.

I would argue that lints running after typeck is not the only concern around making shadowing a lint and not a hard error. If, for instance, we empirically determine that shadowing type/lifetime params is not intended in, say, 90% of code that inhibits it, then there's little value in it being a lint and some potential cost: lints do not stop compilation, therefore typeck errors emerging from an unintended shadowing could obscure the lint and thus make reading the compiler output more difficult.