Warning: Breaking API/ABI changes. New optimizations to decouple update loop from number of buffers queued in audio streaming to allow more for more and smaller buffers to be processed in a loop. This also has the benefit for access_data callback buffer sizes not to be dictated by the performance tuning. Note that the Load APIs break and add a new parameter to specify the target number of buffers to queue per update pass. Also, I changed the access_data from ALboolean to ALuint for possible future use since I'm already breaking the API. The motivation for this is for the feature request of allowing for data callbacks on buffer queue to modify the input data right before submission.