[Performance] use "defer" to avoid loading extra (possibly expensive) data in Task Manager

[kdelee]

In the task manager, we make a series of queries such as:

https://github.com/ansible/awx/blob/devel/awx/main/scheduler/task_manager.py#L90-L100

awx/awx/main/scheduler/task_manager.py

Line 104 in adb2b0d

graph_workflow_jobs = [wf for wf in WorkflowJob.objects.filter(status='running')]

awx/awx/main/scheduler/task_manager.py

Line 112 in adb2b0d

return [invsrc for invsrc in InventorySource.objects.filter(inventory_id__in=inventory_ids, update_on_launch=True)]

We do not need all the fields that come back from these queries. I had attempted to cut down on how much data we load in #12045 by using the only directive, but it was later reverted because we determined that because of ways the polymorphic models have secret fields, that this resulted in MORE queries (@AlanCoding helped me find this by hacking the dev environment and using the django debug toolbar. We made the Task Manager only run when we hit an api endpoint in order to get this information.)

An alternative approach would be to use defer https://docs.djangoproject.com/en/4.0/ref/models/querysets/#defer to specifically exclude certain fields.

I had manually searched through the code to find that we reference ['celery_task_id', 'controller_node', 'created', 'execution_node', 'instance_group', 'job_explanation', 'name', 'pk', 'status'] so the work would consist of finding out which fields are on the models that are NOT in that list, and defer those fields.

This came up while looking at #11671

[AlanCoding]

I would categorize this as more performance rather than tech debt. Just making that point now as I'm trying to organize issues.