Bug 211182

While trying to investigate a different bug, I ran the browser with
`rr record --chaos`, which makes it run very slowly and shuffles
thread scheduling to try to make existing race conditions more likely
to show up, also inevitably making the software run very slow.

Doing so I found something strange: the video kept looping even though
it didn't have the `loop` attribute set.

After some debugging I found that MediaPlayer decides if the video has
ended in part by checking `currentMediaTime()` is greater or equal to
the video duration, which was not guaranteed to be the case in
MediaPlayerPrivateGStreamer.

As a consequence of this patch, one new LayoutTest has passed.