Memory optimized branching operation

[simoneb]

The current branching operation caches all rows in order to supply them to each child operation.
I created a multi-threaded branching operation which does not cache the input rows and supplies one element at a time to child operations by iterating the inner collection one element by one and waiting until all children have consumed that element.
This saves memory avoiding OOME for large volumes of input rows but is more CPU intensive due to thread synchronization.

[jalchr]

I suspect this might lead to transaction dead lock for updates/inserts into a database:
I have this config:
new Multi-threadedBranching().
.Left(new InsertRecords())
.Right(new UpdateRecords())

since the process is transactional, and the two operations might run simultaneously, this will be cause one operation to timeout waiting for the second to complete (which it won't until whole process finishes)