Fix the shutdown checker for a Go worker pool

An AI helping you build a Go worker pool insisted that channels are thread-safe, so several goroutines can close the same channel for a clean shutdown. In Go that panics. Your team now replays recorded channel operations through a checker before trusting a shutdown design.

An assistant wrote find_channel_close_violations(op_lines). It is in the editor. It has a bug. Find it and fix it.

Input and output:

• op_lines: channel operations in the order they happened, each written channel|goroutine|op, such as jobs|worker_1|close.
• Return the list of violations, in the order they happen.

Rules the tests check:

• Split each line on |. Skip a line with fewer than 3 parts. Only the first 3 parts are used.
• Trim whitespace from both ends of each part. Skip the line if the channel or the goroutine is empty, or if the op is not exactly send, recv or close.
• Every channel starts open.
• close on an open channel closes it.
• close on a closed channel is a violation, "close_of_closed:<channel>:<goroutine>". The channel stays closed.
• send on a closed channel is a violation, "send_on_closed:<channel>:<goroutine>".
• recv is never a violation, open or closed.