This new system is used for asynchronous decoding of sound data. The
main improvement with this one over the old one is the ability to do
multi-producer, multi-consumer lock-free posting of messages which
means multiple threads can be used to process jobs simultaneously
rather than a single thread processing all jobs serially.

Decoding is inherently serial which means multiple job threads is only
useful when decoding multiple sounds. Each individual sound will be
decoded serially.

Another change with this commit is the ability for applications to
control whether or not the resource manager manages it's own job
threads. This is useful if an application wants to manage the job queue
themselves if, for example, they want to integrate it more closely with
their existing job system.

Public issue: https://github.com/dr-soft/miniaudio/issues/177

Public issue: https://github.com/dr-soft/miniaudio/issues/177

Public issue https://github.com/dr-soft/miniaudio/issues/176