Investigate possible reconciler refactor to simplify code and improve testing in gitlab-agent

MR: Pending

Description

The reconciler class in the agent handles a lot of complex logic and has a single entry point method. This may be warranted given the complexity of its operations, but this also makes tests difficult to write and makes covering code in tests a chore. It is possible we could break the reconciler further apart and have unit tests on the broken dependent objects to reduce the work needed to test the reconciler on both happy/sad paths. Let's look into this.

Acceptance Criteria

• Investigate possible code refactorings in the reconciler object
• If refactor is feasible, draft up a plan to approach refactor