Ensure `proto::UpdateWorktree::removed_entries` doesn't exceed chunk size

This was causing the database to panic because we were trying to remove too
many entries at once.