CoreAudio may not use the ideal buffer size in the presence of sample rate conversion

[bjornroche]

At the moment, core audio asks the device for a buffer size that is the same as the buffer size requested by the user. This makes sense and works well as long as SR conversion is not taking place. In the presence of SR conversion, a sample frame on the device no longer corresponds to a user sample frame. It would be good if this could be adapted in a better way, if possible, as it would reduce the chances of dropouts, and possibly reduce the CPU usage a bit.