Output the audio size in bytes when logging

[kovidgoyal]

Currently, piper logs something like

Real-time factor: 0.24298105233065118 (infer=2.355538764 sec, audio=9.694331065759638 sec)

In addition to this I would like it to log

bytes=XYZ

this eases integration with other software that is sending multiple sentences and reading the output data while keep track of where one sentence ends and the next begins.

Thanks.