Fix[mqb]: drain on closeChannel

[dorjesinpo]

mqbnet::Channel::closeChannel should close channel (asynchronously) only after sending all items unless channel is closed

[hallfox]

How are we guaranteed that we'll be in the IDLE mode for the channel close logic to kick in here?

[dorjesinpo]

The mode can be any after d_isClosing.store(true);
If it is e_BLOCK, the wakeUp(); should wake it up and the mode goes the cycle
e_BLOCK -> e_FLUSH_BUFFER -> e_IDLE -> e_BLOCK

This logic is based on the fact that after wakeUp();, the mode must arrive at e_IDLE where d_isClosing gets picked up (unless resetChannel()/setChannel() happens which means the channel is closed).

[hallfox]

lgtm

[678098]

We can remove explicit values or start them from 0

[678098]

This enum value is still mentioned in the component's docs.
These docs are also outdated:

blazingmq/src/groups/mqb/mqbnet/mqbnet_channel.h

Lines 48 to 67 in 34422aa

	// Internally, 'mqbnet::Channel' starts a new thread and unconditionally
	// buffers everything it needs to write into single consumer queue. The thread
	// reads the channel state which can be
	// - e_INITIAL, not connected
	// - e_RESET, indicating a need to reset because of a connection change;
	// - e_READY, low-watermark
	// - e_HWM, high-watermark
	// If the state is 'e_RESET', the thread clear the queue and transitions to
	// 'e_INITIAL' . If the channel is set and the state is e_INITIAL, it
	// transitions to 'e_READY'. It the state is not e_READY, the thread waits on
	// condition variable until the state is e_READY. If the state is 'e_READY',
	// the thread pop an Item from the queue and attempts to write.
	// Three public methods can change the state:
	// - 'resetChannel' sets the state to 'e_RESET';
	// - 'setChannel' sets the state to 'e_RESET' and waits for the internal
	// thread to transition to 'e_READY'. This is done to
	// make sure, any write succeeds after 'setChannel'
	// returns. The transition 'e_RESET' -> 'e_READY' is
	// signaled by conditional variable.
	// - 'onWatermark' sets the state to 'e_READY' on LWM and 'e_HWM' on HWM.

[dorjesinpo]

(unrelated) IT failure is analyzed and the fix is created #1257