Rewrite websockets

[juanmc2005]

This PR addresses issue #113

Changelog

• Re-implement WebSocketAudioSource from scratch using websocket-client and websocket-server instead of websockets
• Add diart.client script to run a websocket client that sends audio (from the microphone or from a file) to a websocket server
• Add diart.serve script to run a websocket server that receives audio from a websocket and sends responses in RTTM format
• Move all diart.* CLI tools to diart.console (still accessible as e.g. diart.stream)
• Add convenient diart.utils.encode_audio() and diart.utils.decode_audio() to send and receive audio between server and client
• Include diart.serve and diart.client in websocket section of the README
• Minor cleaning in diart.inference and diart.optim

The goal of diart.client in particular is to be able to test both the default and custom websocket servers easily