This makes ma_decoder_read_pcm_frames() consistent with other data
sources. It now returns a result code and outputs the number of frames
read via an output parameter.

This is now consistent with all other data sources:

  * Returns a result code
  * Length is returned via an output parameter

Use the `encodingFormat` variable in the config instead.

This also renames the `resourceFormat` variables used by the encoder to
`encodingFormat`.

This was used with the old logging level system which is no longer
being used.

This has been replaced with MA_LOG_LEVEL_DEBUG.

This was specific to devices, but now that miniaudio has expanded it's
no longer useful. The new logging system is more generic and useable by
parts outside of devices and contexts.

This has proven to be exceptionally un-useful in practice and the
maintenance cost just isn't worth it.

This might come back later, but for now this is no longer a thing.

This now takes an input and output buffer rather than performing the
clipping in-place.

The previous version only allowed in-place clipping, whereas the new
one supports clipping into a separate buffer. The input and output
buffers can point to the same buffer in which case the clip will be
performed in-place.

This has been replaced with ma_silence_pcm_frames().

This removes the `pBufferOut` parameter from the following APIs:

  * ma_rb_commit_read
  * ma_rb_commit_write
  * ma_pcm_rb_commit_read
  * ma_pcm_rb_commit_write

These are replaced with ma_standard_sample_rate_min/max.

This fixes an issue where excess channels are not heard when the sound
source has more channels than the listener.

This commit also fixes a bug where panning is incorrectly getting
applied to non-directional channels, such as mono and LFE channels.

This is work towards using the heap for storing per-channel data so we
can get rid of upper channel count limit and remove MA_MAX_CHANNELS or
at the very least stop MA_MAX_CHANNELS from affecting overall memory
usage.

Previously this was restricted to 2 input buses and 2 output buses, but
this has been lifted to 254. When the number exceeds 2, internal data
structures will be allocated on the heap, otherwise they'll use a local
array contained within the ma_node structure.

This commit changes the node configuration. Previously there was a
fixed sized array for specifying the channel counts for each bus. This
array must now be defined outside of the config by the caller. The
following config variables have been renamed:

  * inputChannels > pInputChannels
  * outputChannels > pOutputChannels

This commit also adds the ability to configure input and output bus
counts on a per-instance basis rather than via the node vtable. To do
this, set the bus count in the vtable to MA_NODE_BUS_COUNT_UNKNOWN.
This will tell miniaudio to look at the node config to determine the
bus count rather than the vtable. It's an error to specify this in the
node config if the vtable specifies anything other than
MA_NODE_BUS_COUNT_UNKNOWN.