Add MediaSession runtime flag along with stub MediaSession object

LayoutTests/virtual/mediasession/media/mediasession/mediasession-constructor-expected.txt

Index: LayoutTests/virtual/mediasession/media/mediasession/mediasession-constructor-expected.txt
diff --git a/LayoutTests/virtual/mediasession/media/mediasession/mediasession-constructor-expected.txt b/LayoutTests/virtual/mediasession/media/mediasession/mediasession-constructor-expected.txt
new file mode 100644
index 0000000000000000000000000000000000000000..adf19a96ab593a44b19e6c0ee622074966e35302
--- /dev/null
+++ b/LayoutTests/virtual/mediasession/media/mediasession/mediasession-constructor-expected.txt
@@ -0,0 +1,4 @@
+This is a testharness.js-based test.
+PASS Test that the media session constructor exists 
+Harness: the test ran to completion.

[mlamouri (slow - plz ping), 2015/06/29 10:28:31]

I don't think you need expected.txt files for passing testharness tests.

[davve, 2015/06/29 10:45:52]

The reason I put this here is because otherwise it fallbacks to the non-virtual
-expected.txt, which obviously doesn't match (since the flag isn't enabled
there).

I mentioned this in the closed https://codereview.chromium.org/1208963002/ but
failed to bring over the information here.

If you know of a way to avoid this, I'm all ears.

[mlamouri (slow - plz ping), 2015/06/29 10:52:39]

Hmm, indeed, it might be tricky. Is the presubmit hook complaining about this
file? I guess I'm fine as long as the presubmit hook is fine ;)

[davve, 2015/06/29 11:16:10]

The presubmit hook is indeed complaining as I allude to in
https://codereview.chromium.org/1208963002/ (part of why I did the split
initially). This CL is therefore uploaded using --bypass-hooks but I have tried
to eliminate other errors by temporarily removing the -expected.txt file. It's
not a bulletproof procedure though.

Should I file a bug on that (where?) to find out the best course of action?

+