Unify use of 'output' and 'desired' in audio code-base

[NicolasHug]

This PR renames a few stuff in our C++ code. (This isn't about Python code or Python conventions!)

I've been a bit slopy with the use of the desired and output prefixes in the audio C++ code, and this is becoming more evident as I'm working now on sample rate conversions (PR to come). I want to establish a clearer convention for the C++ audio code:

• "desired" is for stuff that are desired by the user, e.g. desiredNumChannels. In C++ we should just represent this as fields in AudioStreamOptions (will create follow-up PR). It usually is an std::optional. When these fields are in AudioStreamOptions (as they should and will), we can ommit the desired prefix.
• "output", as in "outNumChannels" are for properties of the actual output. It can e.g. refer to the number of channels of the decoded tensor (when decoding), or to the number of channels of the encoded data (when encoding). outputStuff is never an std::optional and it is typically set to desiredStuff.value_or(value_of_the_source).

[scotts]

Thank you for making these even more clear!