Add speech recognition feature to web frontend

[joelbudu]

It should support the following

1. Audio upload
2. Record from the browser
3. Realtime speech recognition (if feasible)