Enable per message deflate on server

[pradeepvairamani]

permessage-deflate extension enables the client and server to negotiate a compression algorithm and its parameters, and then selectively applies it to the data payloads of each WebSocket message.

permessage-deflate header is used in the handshake to indicate whether a connection should use compression.

1. When a client sends a websocket request, it send's permessage-deflate in the websocket extensions header if the client browser supports it. The server knows if a client supports compression based on this header.
2. If the server decides to use compression, it responds with the same header similar to an ACK message. The client after receiving the response decides whether or not to compress the data based on the server's response.

Verified that the appropriate headers are being sent back:

By default, the threshold for compression is set to 1024 (in bytes). Messages with size less than this will not be compressed.

[hedasilv]

Is it possible that this setting can cause a spike in memory as we suspected last time we turned it off? If yes, do we want to make it configurable through config.json, so it's easier to disable that if needed during an emergency?

[tanviraumi]

I have the same concern. There can be a CPU and memory spike (gorilla/websocket#203).

@GaryWilber Does push enable it?

[pradeepvairamani]

Yes, I saw Henrique's comment after merging the code. Will add the config as he suggested. Push is not enabling it because AFD does not support this extension. With respect to the memory and cpu consumption, pasting the offline conversation I had with Henrique:

It is a valid point but I do not think we will get affected by it for a few reasons:

The theory that this flag was causing the memory issues is probably not true. The default for this flag was always false..I just added this line explicitly to make doubly sure.

This comes into play only for socket payloads that are more than 1mb..that that does not happen in most cases for us

I stress tested it pretty extensively and did not see any difference in the memory or cpu usage