fix: Re-introduce 1 second sleep to reconcile informer

[simster7]

Fixes: #3665

This line was removed in v2.9.0, but issues arose with the controller picking up outdated versions of workflows, operating on them, then attempting to update the resources on the cluster. Since the workflows were outdated, conflicts arose when attempting to update, causing our conflict resolution code to be invoked. The conflict resolution code was not perfect and workflow information was not correctly persisted, causing undefined behavior.

Testing procedure:

1. Make a build without this line.
2. Run multiple workflows
3. See undefined behavior (hanging pending pods, hanging running nodes, etc.); see the following log:

   Error updating workflow: Operation cannot be fulfilled on workflows.argoproj.io \""WF_NAME\"": the object has been modified; please apply your changes to the latest version and try again Conflict"
   

4. Make a build containing this line.
5. Run multiple workflows
6. See no undefined behavior; don't see the problematic log.

[alexec]

This is the right solution, we should choose reliability over speed every time.

Backport to v2.9 and v2.10?

[alexec]

I think we should look at alternative methods to do this, but not today. Today, lets fix the bug with a know working solution.

[alexec]

This is available in https://github.com/argoproj/argo/releases/tag/v2.10.0-rc6

[alexec]

Available in v2.9.5