[DO NOT MERGE] deny array_into_iter by default (for crater run)

[LukasKalbertodt]

In #66017 I had the idea of letting crater test all regressions from #65819 with the array_into_iter lint on deny. That way we would be able to find out if the lint catches all of the regressions. There is certainly code that could break with #65819 but isn't caught by the lint (see this comment), but we don't know if strange code like that exists in the real world at all.

However, I assumed that we can tell crater to compile the exact version of all crates as in the previous run. Looking at this comment it seems like we can only pass crate names (without version) to crater? So....

• Can we tell crater which versions of crates to build? No
  • Can we even tell crater something like "never use dependency versions released after date X"? No
• If we can't, should we run crater anyway since it's only a few thousand crates?
• Or is there a better we can test the coverage of the lint?

[rust-highfive]

r? @davidtwco

(rust_highfive has picked a reviewer for you, use r? to override)

[joelpalmer]

Ping from Triage: @davidtwco @LukasKalbertodt any updates?

[LukasKalbertodt]

From my side this is ready. We just have to decide if a crater run is useful and then someone needs to prepare a list of all crates that we want to test again and start crater. In my top post I list a few questions about crater; if someone can answer them, that would be helpful.

[davidtwco]

I'm not sure I'm the best person to review this, r? @Centril

[Centril]

cc @pietroalbini for the crater questions.
r? @withoutboats for the libs policy stuff

[pietroalbini]

• Can we tell crater which versions of crates to build?
  • Can we even tell crater something like "never use dependency versions released after date X"?

No.

[LukasKalbertodt]

Given that we can't test for specific versions, I will just close this PR now. I doubt testing "roughly the same set of crates" gives us any useful information.