fix(ext/fetch): retry some http/2 errors

[seanmonstar]

This brings some of the HTTP/2 retry behavior from reqwest to ext/fetch. It will retry very specific HTTP/2 errors once, if the body is able to be used again.

cc #27332

[seanmonstar]

I can't remember, is that CI job saying each PR should have the Cargo.lock completely updated?

[lucacasonato]

I don't know what is up with that. @bartlomieju any ideas?

[nathanwhit]

I believe it just means that running cargo clippy resulted in the lockfile being updated (we run with --locked in CI, so it fails if the lockfile would change). So you should just need to run that locally and commit the updated lockfile.