[TTS] Text preprocessing - remove special characters before sending to TTS engine

[Ironship]

Problem

When generating TTS audio, special characters (HTML tags, markdown symbols, brackets, asterisks, etc.) are sent directly to the TTS engine, causing audio artifacts and mispronunciations.

Current behavior

Only UnbreakLine() and JSON encoding are applied before sending text to TTS engines (in TtsDownloadService.cs).

Proposed solution

Add a configurable TtsTextPreprocessor that:

• Strips HTML tags (<i>, <b>, <font>, etc.)
• Removes markdown formatting (*, **, #, etc.)
• Removes brackets and their content [music], (laughing)
• Strips non-pronounceable characters
• Optionally converts numbers to words
• Configurable per-engine in SeVideoTextToSpeech settings

Files affected

• src/UI/Logic/Download/TtsDownloadService.cs - add preprocessing call
• New: TtsTextPreprocessor.cs
• src/UI/Logic/Config/SeVideoTextToSpeech.cs - add settings

Related upstream issues

• AllTalk TTS: Czech diacritics lost (non-UTF8 form encoding in request) #10133 (AllTalk Czech diacritics lost - related encoding issue)

---

Working on this: @Ironship