Renew slot cache if ClusterPipeline.sync results in MOVED response

[killergerbah]

The intent of this change is to fix a couple issues MultiNodePipelineBase:

• Exceptions thrown from Connection.getMany are swallowed.
• JedisMovedDataException does not trigger a call to JedisClusterInfoCache.renewClusterSlots (and there may be no other opportunity to call JedisClusterInfoCache.renewClusterSlots if the application did not specify a topologyRefreshPeriod when constructing JedisCluster).

With this implementation only the first exception thrown from Connection.getMany is rethrown; however this matches the behavior of Pipeline.

[sazzad16]

Reacting to MOVED response is reasonable request.

BUT the priority is put more towards ensuring execution of pipeline in as many connections as possible.
The finally { executorService.shutdownNow(); } goes against that.

Moreover, there are many cases to consider and I have concerns about those. One simple to understand of those is syncing = false; should be within finally. Because I have concerns overall with the approach, I am not putting it as code review.

[killergerbah]

Got it - I suspect there is anyway a more complete fix for redirect responses from pipelines used on clusters.