Track dictation submit source for autosend

[qqrm]

Motivation

• Prevent automatic sending after a dictation complete event when the dictation was just submitted via the hotkey, by distinguishing hotkey vs mouse submit sources.

Description

• Add timestamp tracking variables lastDictationSubmitViaHotkeyAt and lastDictationSubmitViaMouseAt in src/features/dictationAutoSend.ts.
• Record lastDictationSubmitViaHotkeyAt in the handleKeyDown hotkey submit branch and lastDictationSubmitViaMouseAt in the handleClick submit-click branch.
• In handleTranscribeMessage, skip the transcribe auto-send when a recent hotkey-based dictation submit was the last submit action (checks timestamps and a 1200ms threshold).
• Bump extension version in manifest.json from 1.0.109 to 1.0.110 to reflect the logic change.

Testing

• Ran npm run lint and it passed.
• Ran npm run format:check and it passed.
• Ran npm run typecheck (tsc --noEmit) and it passed.
• Ran unit tests with npm run test and vitest (1 test file) and all tests passed.
• Ran npm run verify (format/lint/typecheck/build checks) and it completed successfully.

---

Codex Task