Remaining MVP TODOs

@here the merge is done, if you want to help

correct README (context: the readme seems wrong, I had to cd server; python app.py, also you need to pip install -r requirements.txt in server/ if you don't have flask)
remove extraneous right arrow (context: there's an extra right arrow that doesn't seem to do anything)
make player opaque and the width of the screen (context: the transparent overlay is just a distraction, should probably be opaque and the width of the screen)
users should probably be able to re-record sounds; it's not useful to fully disable them
- for allowing users to re-record, need to use unique filenames (take uuid out of filename, add a counter)
when you start recording, it needs to pause the sample playback
- stop the sample playback when user begins recording
- pause the sample playback when user begins recording (requires non-native HTML player) (punting on this)
"uh" and "oh" should be separate noises (might need to record a new sound)
and we should clean up the examples on some of them, like
- smooch has double-smooching which will confuse the training, and
- each noise should just have one example sound file imo

general ui cleanup
simplifying the interstitial text
and adding some kind of license agreement to the interstitial should be it and we're ready to publish
stop recording should maybe be in a different spot from start recording, because users will be likely to have talon running with the noise recognizers