Socket behaviour differs from core sockets

[fasteddy516]

I was trying to get the httpserver_websocket.py example from adafruit_httpserver working with a wired WIZnet5K connection and ran into a couple of issues:

1. The handle_http_requests() task was slow compared to the same code using the builtin wifi on a Pico 2 W.

2. The handle_websocket_requests() task failed with an IndexError exception immediately after the websocket connection was established.
    
   I created a test script (available here) that will use the W5K driver if it is available or the builtin WiFi in CircuitPython if it isn't, ran it on a Pico 2 W and a WIZnet W5500-EVB-Pico2 and compared the results. It looks like the non-blocking behaviour defined in adafruit_wiznet5k_socketpool differs from the core socketpool module. adafruit_httpserver expects non-blocking sockets to raise OSError(errno.EAGAIN) when there is nothing to process but the W5K driver returns 0 from recv_into() and ("0.0.0.0", 0) from accept().

I made some quick changes to make the behaviour match, and it seems to make everything work correctly. I can create a pull request, but wasn't sure what other testing should be done to ensure there aren't other unintentional side-effects of the changes.

[justmobilize]

@fasteddy516 I think creating a pull request is good. That way we can comment and test. There are a bunch of use that will run it through it's paces.

[fasteddy516]

@justmobilize Got it. I've found some additional behavioural differences related to the accept() method, though, so I'll try to get those sorted and put it all into one pull request.