Use smarter persistence context for findEach/iterate/stream queries - improve performance of un-tuned findEach/iterate/stream queries that invoke lazy loading

[rbygrave]

These changes should improve the performance of untuned findEach/iterate/stream queries that invoke lazy loading.

Changes:

• Enable use existing persistence context (e.g. beans loaded by prior queries in the same transaction)

• Use a smarter persistence context with a "per 1000 beans added" reset limit. This improves our use of batch lazy loading during iteration

Current limitations

The current implementation of has a few limitations which this change addresses. The findEach etc queries create their own persistence context effectively per "top level bean". The short lived persistence contexts allows these queries to support iterating of many thousands or millions of rows without having to how all the loaded beans in memory (ie. as we no longer have all beans referenced in a single persistence context).