The conversational_speech::Timing class models a list of turns.
Each turn, is identified by a speaker, the audiotrack name, and an offset in milliseconds.
The unit test checks that an instance of Timing is correctly populated and that save/reload leads to the same data.
BUG=webrtc:7218
Review-Url: https://codereview.webrtc.org/2750353002
Cr-Commit-Position: refs/heads/master@{#17346}
