Invert speech logic: speak all content with configurable exclusions

[paulpv]

Currently, CopilotTTS only speaks specific elements (markdown containers, paragraphs, headers, list items). This approach requires maintaining a whitelist of elements to speak.

Proposed Change: Invert the logic to speak everything under TaskChat-module__stickableContainer--* by default, then allow users to exclude specific div classes.

Benefits:

• Simplifies code by removing complex element-specific filtering
• Ensures no content is missed
• More maintainable as GitHub's DOM structure changes
• Users control what NOT to hear rather than what TO hear

Implementation:

1. Remove current element-specific extraction logic
2. Extract all text content from TaskChat container
3. Apply exclusion filters based on user settings
4. Keep existing markdown text filtering for better speech quality

Related: This will work with the new verbosity dropdown (Issue #24 )