Abstract out one off operations within dag

In existing test suite, we have a config called "execute_itr_count". When this is set to N for a particular node, out of 50 odd iterations, this node will be executed only on Nth iteration only. 

Use-case: 

we wish to execute clustering node on 10th iteration. but the entire dag needs to be executed for 25 iterations. 

## JIRA info

- Link: https://issues.apache.org/jira/browse/HUDI-1616
- Type: Test
- Epic: https://issues.apache.org/jira/browse/HUDI-3303
- Affects version(s):
- 0.9.0

---

## Comments

27/Mar/22 14:51;xushiyan;[~shivnarayan] is this done? or push to 0.12?;;;

Start by locating the existing test-suite handling for the execute_itr_count configuration and tracing how a DAG node is selected for an iteration. Determine how the one-off operation should be abstracted while preserving execution of the full DAG for all iterations. Done means the clustering-node use case works at the requested iteration without changing the remaining iterations.

Written by the indexing model from the issue text.