More fixes for async currentSrc behavior

[imbcmdth]

Two main fixes here:

1. The player.currentSrc function didn't allow an empty string to be returned from techs - it would always use src in that case
2. In my testing it appears that currentSrc should always be updated by the time any other loadstart listener is triggered which wasn't always the case using only a setTimeout of 0.

The second bug has the potential to cause problems in plugins that rely on currentSrc being properly updated by the time loadstart is emitted. Since events are triggered synchronously, there are cases where the setTimeout is simply "too late" to match the video element's behavior.

These changes bring it much more in-line with what the video element does

As an example, I ran the test @heff created http://jsbin.com/wulomalopu/1/edit?js,console,output (but modified to use video.js) and the output with these changes on Chrome 43 is:

"el created"
""
""
2
"source set"
"http://vjs.zencdn.net/v/oceans.mp4"
""
3
"async"
"http://vjs.zencdn.net/v/oceans.mp4"
"http://vjs.zencdn.net/v/oceans.mp4"
4
"source set again"
"http://vjs.zencdn.net/v/oceans.webm"
"http://vjs.zencdn.net/v/oceans.mp4"
5
"async again"
"http://vjs.zencdn.net/v/oceans.webm"
"http://vjs.zencdn.net/v/oceans.webm"
6
"empty source set"
"file:///Users/jrivera/projects/video.js/tester.html"
"http://vjs.zencdn.net/v/oceans.webm"
7
"loadstart"
"file:///Users/jrivera/projects/video.js/tester.html"
"http://vjs.zencdn.net/v/oceans.webm"
8
"async for empty source"
"file:///Users/jrivera/projects/video.js/tester.html"
"http://vjs.zencdn.net/v/oceans.webm"

..and that is very, very close to the test's output on Chrome.

[gkatsev]

The other option is to just make setting currentSrc synchronous.

[dmlap]

Should this method be _-suffixed?

[dmlap]

@imbcmdth updateCurrentSource feels like it should be marked as internal. Otherwise, LGTM.

[gkatsev]

Closed via 463ba4e