Responder lifetime cannot be infered

[mrene]

This stackoverflow question shows that it's not sufficient to have elided lifetimes on Responder because these aren't correctly infered when the wrapped responder is less restrictive.

The root cause of this is the PhantomData holding 'r and 'o inside Responder. These aren't required because the bounds on the impl blocks are sufficient to provide the correct semantics, and R will already hold these bounds implicitly.

This is gonna be a breaking change, but it should make it easier to use in the long term. That is unless you can think of any cases that this doesn't handle?

[ELD]

I don't know what's going on with Clippy on the nightly compiler, but that looks like a serious false positive. The stable compiler tests pass, though, so I'm fine with that seemingly erroneous failure.

Thanks for fixing the lints. I imagine the workflow didn't run until it was approved since you're a first-time contributor to the repo.

The explanation on the SO post and the PR makes sense to me and the tests all seem to pass. Looks good!

If you wouldn't mind squashing the changes into a single commit, that would be my preference before merging. Thanks for the contribution! Just remembered I could squash and merge myself. Thanks, again!