[FEATURE] Add Voice Input (Hold-to-Talk) Support to VS Code Extension

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

The Claude Code VS Code extension lacks voice input functionality, creating a usability gap compared to other Claude Code environments (desktop app, web version). Users who prefer voice-first interaction, have accessibility needs, or work through complex problems better via spoken language currently cannot use voice input in VS Code. This limits the extension's utility for brainstorming, rapid iteration, and accessibility compliance.

The current workflow requires users to manually type all prompts, even for complex ideas that would be more naturally expressed verbally.

Proposed Solution

Add native hold-to-talk voice input to the VS Code extension. Implementation should:

Allow users to hold a customizable keybinding (e.g., space, ctrl+space) in the prompt input box to activate voice recording
Display visual feedback (e.g., animated mic icon, waveform) while recording
Automatically submit voice input when the key is released
Support the same voice-to-text transcription pipeline used by the desktop app
Integrate with existing voiceEnabled setting in ~/.claude/settings.json

Alternative Solutions

Use the Claude Code desktop app or web version (claude.ai/code) for voice input, but this breaks the VS Code workflow
Rely on OS-level voice input (Windows/macOS speech-to-text) - which is very buggy and troublesome in windows

Priority

High - Significant impact on productivity

Feature Category

Interactive mode (TUI)

Use Case Example

Scenario: Brainstorming with Claude while coding

I'm working on a complex architectural refactor in VS Code
I want to discuss the approach with Claude before implementing
I hold Space in the prompt box and say: "Walk me through a three-step strategy for migrating this service from monolithic to microservices architecture"
Claude immediately receives my voice input, transcribes it, and responds
This is significantly faster than typing a detailed technical question and more natural for iterative design conversations

Additional Context

The keybindings schema already includes voice:pushToTalk action, suggesting partial implementation infrastructure exists
Voice input works in Claude Code desktop and web; VS Code extension should achieve feature parity
Setting "voiceEnabled": true in ~/.claude/settings.json has no effect in the extension (currently ignored)
This aligns with VS Code's commitment to accessibility and modern developer UX