If the message is NULL just continue. Breaking out of the loop is
wrong because it stops the ecm/processing thread and leaves tsdecrypt
in a non working state.

Handle errors in generic camd code instead of protocol code.

This makes do_emm() and do_ecm() take the same parameters. The next
step is to make them take struct camd_msg as single parameter.

This workaround is no longer needed because camd socket is now set
TCP_NODELAY and this fixes OSCAM communication.

This way it works better when many packets are sended. Without
NODELAY, OSCAM sometimes gets packets boundaries wrong especially
when there are lots of packets (test with emm stream of 16000+
packets per minute).

This prevents unnececary warnings. Sometimes OSCAM responds with error
and on the next query everything is OK. So before raising errors flags
try a couple of times.

Use different queues for ECMs and EMMs. This allows to process
ECMs before EMMs, preventing the case where too much incoming
EMMs slow down ECM processing and interfere with getting the code
words on time.

This fixes Sky Italia HD channels. When they send new
code word the one that is about to expire is set to 0
and this causes short decryption failures. Now if code
word is all zeroes it is not set.

Add --cw-warn-time to set how much time to wait before
starting to complain about code words.

This move is in preparation for adding ecm reports.

The libraries are now free software and are hosted on github.com

This is not strictly correct but helps to not spam the logs
with multiple connects per second if camd server is down.