Fixes #405

This approach removes the JobSet status update API calls from the controller, and simply modifies the status in memory, and does a single status update call at the end of each reconciliation attempt.

These changes also include:

• Refactoring to improve readability and organization
• Fixing bug causing unnecessary status update calls (here the replicated job statuses must be sorted before doing the semantic equality test)
• Refactor "in order startup policy" condition into 2 mutually exclusive conditions: one indicating the startup policy condition is in progress, and one indicating it has completed.
• Remove forceFalseUpdate flag from conditionOpts since we are no longer using a condition status of false/true on the in order startup policy condition to indicate if the startup policy is in-progress/completed (see prior bullet point).

Note: the operation conflict errors are still in the integration tests. One idea I had is that we are still doing Job status updates in the middle of reconcile logic. These will trigger Jobset reconciles which then update the jobSet.status.replicatedJobStatuses as it looks at those child jobs. The controller runtime work queue should be handling these sequentially though since MaxConcurrency defaults to 1 and we don't override that, so I'm not sure of the root cause yet. Kueue has these errors in the integration test logs as well though so they may be an expected side effect of the integration test framework or someting.