Add 'FinalStop' hook that fires only once after all processing is complete

Description:

Currently, the Stop hook can fire multiple times during a single Claude response, presumably when there are multiple processing stages. This makes it difficult to trigger actions that should only run once after Claude has completely finished responding.

Current behavior:

• The Stop hook fires "when the main Claude Code agent has finished responding"
• In practice, this can trigger multiple times in a single response

Requested feature:
Add a new hook type (e.g., FinalStop or ResponseComplete) that:

• Fires exactly once per Claude response
• Only triggers after all processing is complete and Claude is waiting for the next user input
• Does not fire on user interrupts (similar to current Stop hook behavior)

Use case:
This would be useful for:

• Playing a notification sound once when Claude finishes
• Running cleanup scripts after complex operations
• Triggering external integrations that should only run once per response
• Logging/analytics that need to capture complete responses