Alerta bridge shows problems with massive alert processing. After some time
Alerta.io is blocking the socket and this leads to Kafka timeouts.
Therefore, an timeout is defined and when the error appears then all successful
alerta.io request before are committed and an exception is thrown. Then Kafka tries again
the current offset but does not repeat the committed offsets before.
The heartbeats are needed to enforce triggering the Alerta filter window in case
there is no other kafka updates.

Signed-off-by: marcel <wagmarcel@web.de>