signature: reconsider `allow(async_fn_in_trait)`?

[tarcieri]

I'm curious why we're suppressing this lint. The reason given in #1419 which did the original AFIT migration is:

#[allow(async_fn_in_trait)] is required until a solution to RPITIT is merged in rust.
This put responsability on the implementor of async_signature::AsyncSigner to make sure their future is Sendable.
see rust-lang/rust#115822 (comment) for more context.

RPITIT shipped in 1.75.0, so should we reconsider this?

cc @baloo