AI TTS as replayable artefact (similar to chat gateways)

This issue is really a combination of a few correlated issues.

Hermes WebUI offers the option to show a speaker button underneath LLM output. On every click, the voice message is regenerated.

• If using an AI provider there's a lag of several seconds. But there's no progress indicator. It happened to me that I was wondering whether anything is happening, clicked it a second time a few seconds apart, and then got double output with voices speaking on top of each other. -> If I click the button and voice generation is happening in the background, the button should be greyed out, and maybe a spinner should be shown in its place to signify something is happening.
• After the text has been read, its gone. On Telegram, the transcript remains as an artifact so I can play it again. -> Hermes WebUI could keep a cache of generated voice artifacts.
• One of the Preferences option says "Text-to-Speech for responses - Show a speaker button on each assistant message to read it aloud using your browser's speech synthesis.", but below there's an option to choose other TTS providers like OpenAI on the server. This is confusing. Similarly, the hands-free option also says that it requires browser STT and TTS support, although hermes has that built-in and I'm not sure if the text is wrong. The options should probably say something like "use the STT/TTS provider configured below". -> The whole STT/TTS section in the settings needs a bit of an overhaul.