Feature: Dynamic Model Download for TTS tab

[jordyantunes]

Hey, I'm creating a PR for adding dynamic model download for the TTS tab. Let me know if you think this is useful for you and if you have any comments/concerns.

What this PR does

• Updates api/pull endpoint to support pulling Kokoro and Piper's models and voices, with optional parameters (allow_patterns and ignore_patterns) to control what files should be downloaded
• Added dynamic model download to TTS page by checking if model files exist, displaying a Download button and using the api/pull endpoint
  • Since downloading all of Piper's voices may take sometime (and we may get rate limited by HuggingFace), I added a radio button with a few option to download specific languages/voices and a Custom Voices field
• Added follow_redirects=True to API calls for endpoints behind ingresses that may redirect

TTS tab working with new changes:

speaches-demo.mp4

[YevheniiSemendiak]

Hey @lilleswing , maybe you had a chance to take a look on this?

It would be really beneficial for the cases, when you host Speaches behind the proxy or when you want to select which voices to load.

[fawzib]

maybe add another feature to also delete