Support custom JSON encoders

[Gelbpunkt]

This adds support to do things like bot.wavelink.set_serializer(ujson.dumps) and pretty much any other JSON encoder that dumps to strings or bytes. I did not know a better way to do this other than 3 attributes on all classes to make sure they are being created correctly for any future cases.

I left a note regarding sending bytes, as it would be technically more efficient, yet Lavalink rejects binary traffic for now. Might want to request that in upstream.

I've tested the basic idea of dumping and then using send_str in production with orjson and the send_json from aiohttp pretty much dumps it and sends it then anyways, so there is no overhead created here, yet using a kwarg to the send_json would not allow JSON encoders that return bytes, so I decided for this way.

[EvieePy]

Thanks for the PR. It has been rebased.

Was just wondering if there are any checks we can do in the set_serializer function to ensure a valid encoder is passed?

[Gelbpunkt]

@EvieePy we could theoretically dump any dictionary, e.g.:

result = self._dumps({'a': 1, 'b': 'hello'})
assert result.decode('utf-8') == '{"a": 1, "b": "hello"}'