Optimize UIMA-AS code that collects pipeline stats

When collecting stats from AE Management objects the UIMA-AS needs to create a snapshot of stats before calling AE process method. It actually does a deep copy (by serialization) of Management objects. This snapshot is used to calculate component analysis time when the process method completes. The problem is that the serialization is adding too much overhead and actually is not necessary. Refactor code to avoid serialization overhead.