My current plan of attack is to flip rtc_include_tests over to true by
default and make Chrome explicitly set it to false. This means I must
land this change first and then the WebRTC change. This won't make
parity with the WebRTC .gyp files worse since they handle this
particular thing different anyway (.gyp files look at
build_with_chromium to make this decision, I believe).

BUG=webrtc:6119

Review-Url: https://codereview.chromium.org/2162573002
Cr-Commit-Position: refs/heads/master@{#406329}