the initial pull request for DSharpPlus.Voice.

known issues:

• currently, it will explode upon receiving RTCP packets.

todo:

• receive
• audio packet queuing for sending
• reconnecting/resuming
• rewire encoding to take advantage of opus' 24-bit and float support instead of (not very costly, but absolutely unreadable) normalization to 16-bit
• user interface for sending/receiving audio
• use a more sophisticated mechanism for audio enqueuing than Channel
• reduce memory management overhead
• handle RTCP packets
• correctly track and send RTCP information
• docs

optional:

• port resampling to managed
• expose received RTCP information to users?

open questions:

• is it relevant that if we don't receive a user's SSRC in time, we can't decrypt their audio? do we have to buffer those frames?
• do we need a whole native dependency (speexdsp, 616kb, for reference - everything else is about 16.5mb combined) for resampling 44.1KHz to 48KHz or does it make more sense to implement that ourselves? - is it feasible for users to use the resampler themselves
• is it fine that if no more audio is available to send, we extrapolate silence, or is that something that can happen in the form of brief user-code-side lag and we should just not send anything? (essentially, is Channel fine as a queuing mechanism or do we need it to be more sophisticated)
• is it fine that every connection has its own PeriodicTimer for timing sends? and - is it fine that it ticks every 20ms, even if opus theoretically supports shorter frames?