A previous survey of our implementation indicated that we do at least one more post onto the renderer main thread than the specs require during attachment of a MediaSource to an HTMLMediaElement and scheduling 'sourceopen' event on the MediaSource.

This bug tracks:
1) adding metrics to better understand the effective latency (at least in perf telemetry, if not also in UMAs/UKMs), and
2) reducing the number of hops back onto the renderer thread before scheduling 'sourceopen'

Initially targeting M-64, though very likely (2), if not also (1) will land mostly in M-65.