Mic button added to chat composer for voice input via speech recognition

Evanfeenstra adds a microphone button to the ChatInput component, letting users dictate prompts instead of typing. The change is frontend-only and keeps everything downstream of the text field unchanged.

The change lands at the chat composer layer. A mic button sits alongside the existing text field; activating it starts speech recognition and transcribes speech into the same input the user would otherwise type into. The underlying chat and model logic is untouched.

The implementation detail the PR doesn't spell out is the recognition mechanism. The standard approach for in-browser mic input is the Web Speech API, which runs client-side with no server round-trip. That would explain the tight scope - no backend changes, no new dependencies visible from the PR description.

One thing to note for teams tracking stakwork/mikeoss as a base: since this merged to main, voice input is now the default chat experience in that fork. Anyone syncing from it gets the change automatically.

The PR description is one line, so there's limited signal on edge cases - fallback behavior when the browser doesn't support speech recognition, or how partial transcripts are handled if recognition is interrupted.