Implement a cache for the TTS

[enrico3]

I find the delay waiting for the TTS a bit too long. Some ideas to solve it:

1. Cache the mp3 locally and play the local file when you already downloaded it. Only delete mp3s at shutdown or with a user option.
2. Additionally download the TTS of the whole text / current page / some words ahead in advance so you always have the local file ready when the user requests it.