[Feature request] Text-to-speech output

Description:
It would be helpful to have a text-to-speech option that reads Claude's responses
aloud. This would improve accessibility and allow users to listen to responses while
multitasking.

Proposed implementation:

• Add a --tts flag or setting to enable text-to-speech
• Integrate with system TTS (espeak on Linux, say on macOS, etc.)
• Option to automatically read responses or trigger on demand

Use case:
Currently there's no easy way to have Claude's responses read aloud without manually
copying text to external TTS tools, which interrupts the workflow.

Alternative considered:
Piping output to external TTS tools, but Claude Code's interactive nature makes this
difficult.