All of the buffer size returned by Windows Core Audio APIs are in unit
of audio frames (which is sample times number of channels), while
WebRTC's AudioDeviceBuffer RequestPlayoutData method takes in samples
per channel (equivalent to frames per channel) but returns number of
audio samples in all the channels. This CL makes sure that we compare
playout block size in frames with frames and size in samples with
samples, which should fix the excessive logging issues and audio quality
problems due to the mismatch when comparing.
BUG=webrtc:7797
Review-Url: https://codereview.webrtc.org/2933953003
Cr-Commit-Position: refs/heads/master@{#18546}