- adds resumability tests for the iterator based API (tryNext, hasNext and next)
- adds resumability tests for the event emitter-based API
- adds tests demonstrating that change streams do not resume on
   unresumable errors for iterator and event emitter based apis
- adds tests for resumability iterator mode on 3.6 servers
- adds resume options tests to resumability tests

This commit removes a custom function we had to wait until the topology
was reconnected in favor of performing server selection.  This is described
here: https://github.com/mongodb/specifications/blob/master/source/change-streams/change-streams.rst#resume-process

…ntemitter modes

	tryNext is not blocking and on sharded clusters we don't have control of when
	the actual change event will be ready on the change stream pipeline. This introduces
	a race condition, where sometimes we receive the change event and sometimes
	we don't when we call tryNext, depending on the timing of the sharded cluster.

	Since we really only care about the resumability, it's enough for this test to throw
	if tryNext ever throws and assert on the number of aggregate events.

…r methods

- removes MongoClient.connect() call from new change stream tests
- awaits a promise in iterateUntilDocumentOrError in the unified runner
- removes an unused method in the change stream class

…ise call

- use to.be.a('number') in assertion
- remove bracket style type assertion in favor of as

…ing"

This reverts commit 50980b8.