avoid incoming breakage of for<'a> &'a T: 'a bounds

[aliemjay]

rust-lang/rust#98109 fixes a soundness hole in rust. Unfortunately, this fix exposes an other known bug in reasoning about bounds of the form for<'a> &'a T: 'a. This bug can be illustrated in the following example which is rejected since v1.0!

fn require_bound<T>()
where
    for<'a> &'a T: 'a,
    // T: 'static // <- rustc incorrectly thinks it must prove this bound for the above bound to hold.
{
}

fn test<T>() {
    require_bound::<T>();
    //^ ERROR  type parameter T doesn't live long enough
}

One way to work around such bug is to simply remove the bound since it is always satisfied for any type T.

We tried to fix this in a timely manner, but it's not a particularly easy problem. Indeed it is part of a long-term initiative under the name of "implied bounds" to better support these bounds. Sorry for any inconvenience!

[Dessix]

Confirmed locally and merged; Thank you for your contribution.

[Dessix]

I'll note that #3 covered a bound missed by this PR on line 241. I'll be partially merging that as well. Following that, I will be publishing an update to a more-recent Rust version, which will set this in stone.