Audio no longer needs to be the same format/sample rate to play correctly. This
should be the last step to making audio work correctly (minus several
optimizations still yet to be done).
For accuracy, whenever we need to perform an audio resample, we need to do it
in advance. This is in a process called "conforming" and this commit introduces
the framework by which the decoder can automatically conform an audio stream
to arbitrary parameters for accurate rendering.
Rather than convoluting the MediaInput node, te MediaInput is now an abstract
base class that provides access to a Decoder and derived classes are
responsible for handling it
itsmattkc