* Update shared subscription test to remove invalid assertions on a property that does not necessarily hold (all subscriptions receive messages)
* Update SelfPubSub test to not use counting variables that are pointless (they won't ever be wrong) as well as unsafe from race conditions (they are set and read from different threads with no explicit synchronizations, also the future that was waited upon strictly happens-before the setting of the counter which leads to consistency issues)
* Update the double client id connection failure test to be more forgiving to potential race conditions between the two clients fighting over the client id
* Adds a delay between subscribe and disconnect in the will test. Did not get an encouraging answer when the service team was queried about whether subscribe completion had any eventual consistency issues relative to a subsequent publish (connection will).


Co-authored-by: Bret Ambrose <bambrose@amazon.com>