(edge-http) Server non-static handler

[ivmarkov]

(
This change is long overdue and I was delaying it in the (now lost) hope that eventually, the server Handler can become compatible with the nightly AsyncFn closures.

Well, it cannot, and there is no way it could, because AsyncFn closures - just like regular ones - cannot be generic over their parameters, which is a road blocker.

If it is not clear, our handler needs to be generic over the socket type T, and I don't see any way around that.
)

Putting aside this ^^^ topic, what this PR does is to move the generics (including the problematic 'b lifetime generic) to the Handler::handle method, instead of having these on the Handler trait itself.

This is necessary, or else - as is the case in the current code - we need at a few places to put for<'t> type constraints - which - rather than having the desired effect of "for any lifetime" are instead seemingly ending up requiring the lifetime to be equal to 'static.

I'm not sure, but I think this might be another occurrence of this problem.

The PR is much noisier as it also contains a change to suppress the Clippy complaints for us using an unnecessary named lifetime parameter 'b where we can get-by with just '_.

[ivmarkov]

@AnthonyGrondin FYI if you are using/implementing the Handler trait, there is an adjustment to its signature that you need to do. Nothing super-major though, and I wonder how the current handler signature is actually not affecting you in the first place?

[AnthonyGrondin]

The current handler situation is affecting me in the sense that I need static lifetimes, on anything that passes to the handler. For example, this currently makes it so that I have to statically borrow peripherals if I run the server in a loop, because it could be borrowed on the previous iteration.

See: esp-rs/esp-mbedtls#46 (comment)

[ivmarkov]

The current handler situation is affecting me in the sense that I need static lifetimes, on anything that passes to the handler. For example, this currently makes it so that I have to statically borrow peripherals if I run the server in a loop, because it could be borrowed on the previous iteration.

See: esp-rs/esp-mbedtls#46 (comment)

Hm, not sure that the existing handler does enforce 'static on anything but the edge-nal traits and thus on the embassy-net types.

But perhaps you can try this branch here if it solves the problem for you? It definitely does for me, as now I can use non-static edge-nal and embassy-net types.

[AnthonyGrondin]

I can confirm this fixes the issues for me. I updated what I've been working on with esp-mbedtls locally, and I can use the server without requiring static lifetimes. I can also mutably borrow in a loop without triggering a lifetime issue.

I've looked at the changes, and it's mostly generics / lifetime changes, and other metaprogramming markers, rather than logic changes so it LGTM.