conn_mgr_tests were occasionally failing when executed under heavy load.
The reason for this are short timeout values used when waiting for
events - if executed under heavy load, the even might not be delivered
on time.
An obvious solution would be to simply increase the timeout, however
when testing it turned out the actual timeout value would need to be
pretty high to ensure tests reliability. Therefore, to avoid
unnecessary increase of the overall test execution time to protect
against rare edge cases, rework the event waiting mechanism to use
semaphore instead. The test cases will now specify the events they're
waiting for, and the event handlers will feed the semaphore if all
expected events have been delivered. This allows the increase the
maximum timeout when waiting for events, w/o affecting too much the test
execution time under normal conditions.
Signed-off-by: Robert Lubos <robert.lubos@nordicsemi.no>
45a79d3208