Push Delete doesn't work for Asynchronous Sync

This creates a circular dependency between salesforce_mapping and salesforce_push.
Need to move this to salesforce_push somehow, with a new event if need be.

+  $queued_items = \Drupal::database()->select('salesforce_push_queue', 'spq')
+    ->fields('spq', ['item_id'])
+    ->condition('entity_id', $entity->id())
+    ->condition('name', $mapping->id())
+    ->condition('op', 'push_delete')
+    ->execute();

Entity + Mapping will give you a single, unique result.
Otherwise, there may be multiple Mappings for a single entity type.
If the Entity gets deleted, then we should iteratively delete all mapped objects and queue records.

+  $mapped_object = $storage->loadByEntity($entity);
+
+  if (!$mapped_object) {
+    return;
+  }
+  // Is there ever more than one here?
+  $mapped_object = current($mapped_object);
+  $mapping = $mapped_object->getMapping();

This looks good with one more minor change:

+      $queued_items = \Drupal::database()
+        ->select('salesforce_push_queue', 'spq')

The queue table gets created JIT, so it might not exist at this point, in which case this will fatal.
If the table doesn't exist, we should skip the for-loop and go straight to deleting the mapped object(s).

We should have some test coverage for this at some point, but can commit without it.

This looks good.
I updated the push queue test to demonstrate this bug, attached.

Now combined patch from #9 to address the failing test.

Dunno what's up with #13 failing - it's passing on my local 2 different ways.
Can someone else try running locally to get some insight?

Thanks for this - committed to 4.x and 3.x