Fix race condition in async exception handling by combining JMS ExceptionListener and TransportListener notifications

[jeanouii]

[profdiegocbcastro]

This looks like a race between the transport layer exception
notification and the JMS ExceptionListener callback.

Since TransportListener callbacks may be triggered by the
network transport thread while ExceptionListener notifications
are handled in the JMS client layer, both can fire concurrently
when a connection failure occurs.

If both handlers attempt to react to the same failure
(e.g. closing the connection or restarting consumers),
the timing difference can lead to inconsistent states or
missed notifications.

One possible approach would be to funnel both events through
a single exception dispatch mechanism inside ActiveMQConnection,
so that:

TransportListener.onException()
↓
central exception handler
↓
ExceptionListener.onException()

This would ensure ordering and prevent concurrent handling
of the same connection failure event.