[FEATURE] Make voice dictation timeouts configurable (15s silence auto-stop, 2min max recording)

Feature description

Voice dictation in the Claude Code CLI currently has two hard-coded limits (per https://code.claude.com/docs/en/voice-dictation.md):

• Recording auto-stops after 15 seconds of silence
• Recording ends after 2 minutes total

Neither limit is configurable — there is no settings.json key and no environment variable for them.

Problem

When dictating longer prompts (describing a task, thinking out loud, formulating a multi-part request), both limits get in the way:

• 15 seconds of silence is short when you pause to think mid-sentence — the recording cuts off and you have to restart dictation and stitch the input together.
• 2 minutes total is quickly reached for anything beyond a short instruction.

Proposed solution

Make both timeouts user-configurable, e.g. in settings.json:

{
  "voice": {
    "silenceTimeout": "30s",
    "maxRecordingDuration": "5m"
  }
}

(Exact naming/nesting up to you — env vars would also work.) Reasonable upper bounds are fine; the point is that the current defaults are too tight for longer dictation and cannot be adjusted at all.

Context

• Related but different: #30761 asks for a continuous listening mode; this request is just about making the existing limits configurable.
• Platform: macOS, Claude Code CLI (TUI)