Add RemovableChannelHandler conformance to MessageToByteHandler

[AlexisQapa]

Add RemovableChannelHandler conformance to MessageToByteHandler as a follow up for #1991

Motivation:

RemovableChannelHandler conformance was missing MessageToByteHandler while ByteToMessageHandler did conform to it.

Modifications:

Add RemovableChannelHandler conformance to ByteToMessageHandler
Use the same pattern as for ByteToMessage with hooks for encoders to act on removal.
Added hooks have default implementation.

func encodeLast(data: OutboundIn, out: inout ByteBuffer) throws
mutating func encoderRemoved()
mutating func encoderAdded()

edit: removed context from hooks

Result:

MessageToByteHandler now conform to RemovableChannelHandler.

[swift-server-bot]

Can one of the admins verify this patch?

[swift-server-bot]

Can one of the admins verify this patch?

[swift-server-bot]

Can one of the admins verify this patch?

[swift-server-bot]

Can one of the admins verify this patch?

[swift-server-bot]

Can one of the admins verify this patch?

[swift-server-bot]

Can one of the admins verify this patch?

[swift-server-bot]

Can one of the admins verify this patch?

[swift-server-bot]

Can one of the admins verify this patch?

[swift-server-bot]

Can one of the admins verify this patch?

[swift-server-bot]

Can one of the admins verify this patch?

[swift-server-bot]

Can one of the admins verify this patch?

[swift-server-bot]

Can one of the admins verify this patch?

[AlexisQapa]

I must say I'm not sure about my implementation as I'm quite new to the code base. I'll look into tests when I can thought any advice about how to would be appreciated. Should this be a draft PR ?

[AlexisQapa]

I could turn the RemovalState to a Bool but I'm unsure if it then needs a precondition in handlerAdded.

https://github.com/AlexisQapa/swift-nio/tree/use-boolean-instead-of-removal-state

[Lukasa]

You shouldn't need the precondition in handlerAdded.

[Lukasa]

Great, this looks really good! One final note: can you add a few unit tests that test the paths through the removal flow here? That way we won't accidentally regress this behaviour.