Add REDUNDANT_IMPORTS lint for new redundant import detection

[compiler-errors]

Defaults to Allow for now. Stacked on #123744 to avoid merge conflict, but much easier to review all as one.

r? petrochenkov

[compiler-errors]

nominating for T-lang to see if we'd like to keep this as allow or make it warn, etc., and perhaps to follow-up on the policy.

for the record, I think this lint is perhaps noisy but also very useful, so I'd like to see it at Warn level eventually. just not curious what the best rollout strategy is for that.

[traviscross]

@rustbot labels -I-lang-nominated

We discussed this in the lang meeting today. We came to the consensus that we want to split out into a separate allow-by-default lint (or, alternatively, not lint at all) the case of an import that is redundant due to a glob import.

There are two reasons for this.

One, and most importantly, is that if we were to not do this, a warn-by-default lint here would make adding functions, statics, consts, etc. to a prelude (including the one in the standard library) so disruptive as to have to be treated almost like a breaking change.

Two is that not doing this would make it annoying to migrate away from a glob import incrementally.

Once this case is separated out, we're OK with the remainder of this lint going directly to warn-by-default.

[petrochenkov]

Let's first land the renaming and the warn -> allow change, and then change the lint logic itself.
@rustbot author

[compiler-errors]

OK, I fixed the one nit. Haven't split/suppressed the glob behavior. I can open an issue to do that later.

@rustbot ready

[petrochenkov]

r=me with #123813 (comment) addressed.
@rustbot author

[compiler-errors]

What the heck, I thought I literally fixed that. Sorry haha

[petrochenkov]

@bors r+

[bors]

📌 Commit f6f587e has been approved by petrochenkov

It is now in the queue for this repository.