feat(mobile): add native mic transcription

[0bkevin]

Summary

• Add native and Web Speech API dictation for editable Hermes prompts.
• Coordinate speech recognition ownership across composers and handle lifecycle interruptions.
• Add permissions, locale normalization, transcript merging, error messaging, and composer UI states.
• Document development-build requirements and add unit tests for speech recognition behavior.

Testing

• Not run; change request content generated from the provided diff.