do not panic during chunk dispatching if consumer suddenly closed

[rsperl]

We've seen repeated panics when a consumer suddenly goes away:

panic: send on closed channel

goroutine 605 [running]:
github.com/rabbitmq/rabbitmq-stream-go-client/pkg/stream.(*Client).handleDeliver(0x140004d4000, 0x1400099c000)
	/root/go/pkg/mod/github.com/rabbitmq/rabbitmq-stream-go-client@v1.5.3/pkg/stream/server_frame.go:410 +0xaa0
github.com/rabbitmq/rabbitmq-stream-go-client/pkg/stream.(*Client).handleResponse(0x140004d4000)
	/root/go/pkg/mod/github.com/rabbitmq/rabbitmq-stream-go-client@v1.5.3/pkg/stream/server_frame.go:84 +0x294
created by github.com/rabbitmq/rabbitmq-stream-go-client/pkg/stream.(*Client).connect in goroutine 601
	/root/go/pkg/mod/github.com/rabbitmq/rabbitmq-stream-go-client@v1.5.3/pkg/stream/client.go:196 +0x86c

This PR ensures that the consumer.chunkForConsumer channel is open before dispatching a chunk to it.

[Gsantomaggio]

Thank you, @rsperl. That's interesting. I had never encountered this condition before, and the fix makes sense to me.
@hiimjako WDYT?

[hiimjako]

It never happened to me too. It is strange to me because the client is set to close before closing the consumer.chunkForConsumer, so it should not happen.
But anyway, it makes sense to me too.

[Gsantomaggio]

@rsperl
Is there a way to reproduce this problem?
Did you have this problem? If yes, can you please add more info?
Thank you

[rsperl]

@rsperl Is there a way to reproduce this problem? Did you have this problem? If yes, can you please add more info? Thank you

Unfortunately, i don't have an isolated example, only a very complex one in code I cannot share, though it was fairly reproducable. We have a rabbitmq proxy server using this library, and when we have multiple consumers processing many events, suddenly stopping the consumers (ctrl-c or a kill -9) will cause a panic in the proxy server with the stack trace above.

[Gsantomaggio]

We have a rabbitmq proxy server using this library, and when we have multiple consumers processing many events, suddenly stopping the consumers

It seems an interesting use case!

Out of curiosity: How is it going with the library (except for this bug)?

[rsperl]

We have a rabbitmq proxy server using this library, and when we have multiple consumers processing many events, suddenly stopping the consumers

It seems an interesting use case!

Out of curiosity: How is it going with the library (except for this bug)?

I don't, but we only started using streams a couple of months ago (and we are using rabbitmq 3.12 btw).