The integration test that checks that an orderer is evicted from a channel
and stops its service for the channel has broken due to:

1) A removed log message that it used as an indicator was removed
   in a parallel CR.
2) In another parallel CR, the communication layer now puts message into
   the log asynchronously and doesn't block - and as a result -
   a node might be evicted from the channel but the other nodes will
   close the connection to it before it has a chance of obtaining the block
   that evicts it from the channel.

For (1) - the message that no longer exists was removed from the test.

For (2) - the node that is removed is now always the leader, and this way
          it always gets the block update (because it sends it in the first
          place).

Change-Id: Ib67d1a448447ef44d9b41f52c8ee8bddb6b064ce
Signed-off-by: yacovm <yacovm@il.ibm.com>