Rework

[dryajov]

This is a minimal functional implementation that reworks:

• API, it adds two methods to receive data
  • proc recv*(ws: WebSocket, data: pointer, size: int): Future[int] {.async.}, allows passing a pointer to up to size bytes, note that this will potentially consume multiple and/or partial frames. This will consume a frame as soon as it arrives, without buffering, which allows streaming.
  • proc recv*(ws: WebSocket, size = WSMaxMessageSize): Future[seq[byte]] {.async.} a convenience method that accumulates frames up to a max message size
  • Note that in WS messages split into arbitrary frames delimited by a fin flag
• Fixes an important issue with framing, where a message would be stuffed into one frame, encoded and then split into chunks of 1MB, this is dangerous as anything else writing to the same pipe in between chunks will get interleaved with the frame chunks (this issue is also present in the nim-news package!), the correct behaviour is to split the data into smaller chunks and then frame that
• Fixed control message handling
• Fixed closing flow handling (NOTE, Websockets do not support half-closed/open flow)
• General cleanup and testing

This PR does not address any of the HTTP handling functionality already present, which is largely based on the nim-json-rpc implementation but should potentially be reworked with the new HTTP functionality recently added to chronos.

[sinkingsugar]

Well that was a rework indeed.
LGTM, maybe adding coverage tracking would help as it definitely needs lot of more testing to find runtime issues.

[arijitAD]

typo: greater

[dryajov]

This should be ready for review. As it is, it still lacks comprehensive tests, but it is a fully functional implementation with browser interop.

The next steps are:

• HTTP rework and SSL support
• Adding more comprehensive tests
• Adding browser centric tests perhaps with karma or something similar
  • the most important point is that browser interop is guaranteed

[dryajov]

CI is broken for i386 for github actions here is an issue tracking it - actions/runner-images#2919 (comment). The workaround didn't work, but it's not critical to merge.