This implements atomic processing service
that is able in bulk to update multiple builds,
stages and pipelines.

This uses in-memory structure to calculate statuses
and uses an efficient fetch of DAG dependencies
for efficient processing

This still performs some redundant operations,
but overall it should be significantly faster
processing pipelines of 1000 builds, with
multiple DAG dependencies, as it removes
a number of N+1 problems present in previous
implementation.

This also is order-independent, so it is able
to easily reschedule itself once needed.

This code is also concurrent safe, as if needed
the further processing will be re-scheduled
that will resolve any conflicts.