Remove workarounds for Rust bugs in Yoke, DataPayload

[sffc]

There are docs tests in the yoke and icu_provider crate that will begin compiling in Rust 1.56 or 1.57, due to fixes to upstream bugs rust-lang/rust#86703 (fixed in 1.56) and rust-lang/rust#88446 (fixed in 1.57). This issue is to track steps to update to these Rust versions once stabilized and at the same time update our docs tests.

As part of this bug, also remove the obsolete with_capture variants of the functions, and fix alignment between Yoke and DataPayload.

CC @Manishearth

[sffc]

This is partially fixed by #1090 and #1135, but we are now blocked by the following bugs:

• for<'a> Fn(<A as B<'a>>::C) fails incosistently and requires unnecessary type annotations rust-lang/rust#90875 (_with_capture)
• HRTB bounds not resolving correctly (take 2) rust-lang/rust#89196 (YokeTraitHack)

This new state is reflected in #1134. Moving to the 0.5 milestone to revisit.

[Manishearth]

There may be a workaround that means we can get rid of _with_capture: rust-lang/rust#89436 (comment) (but it requires more annotations)

[sffc]

I'll take another look now that 1.58 is stable.

[sffc]

OK, I was able to enable all of the docs tests that we had marked as "to be enabled" after 1.57.

It appears that we still need YokeTraitHack and attach_to_cart_badly. However, I'm not sure how to test whether we still need project_with_capture; it takes a fn argument, and I assume we'd need to test against a function signature that takes a function trait argument.

Many of the call sites reference rust-lang/rust#84937, which is closed, but the code still doesn't compile (perhaps for different reasons than the issue was originally raised for). I saw this comment, but I'm not immediately sure how to apply it.

rust-lang/rust#89436 (comment)

I'm going to reassign this issue to Manish to take another look at some point before we stamp 1.0.

[robertbastian]

1.61 seems to make attach_to_cart_badly unnecessary, its docs test starts compiling.

[Manishearth]

Let's keep it around but start migrating

[sffc]

Discuss: Should we keep the badly methods or not?

[sffc]

Discussion:

• @Manishearth - For _badly, we should keep them until at least 6 bug-free months have passed.
• @sffc - How about _with_capture?
• @robertbastian - Can we upstream our tests to the Rust compiler?
• @Manishearth - We should try to upload minimal working examples, which might be tricky, but yes, I would support that.
• @Manishearth - We should add the buggy version of map_project so that we can test it.

AIs:

• @Manishearth to file an issue to track adding the tests.
• @Manishearth to change map_project and map_project_cloned to use a closure.

[sffc]

This is 1.0 because it affects DataPayload APIs.

[Manishearth]

Upstream for map_project: rust-lang/rust#99257