Add better defensive checks and error tracking on exercise publishing

[bjester]

Observed behavior

The following issue occurred publishing a channel on production. Although, we never received a report of the issue in our Sentry monitoring:

Traceback (most recent call last):
  File \"/contentcuration/contentcuration/contentcuration/viewsets/base.py\", line 920, in create_change_tracker
    yield tracker
  File \"/contentcuration/contentcuration/contentcuration/viewsets/channel.py\", line 490, in publish
    channel = publish_channel(
  File \"/usr/local/lib/python3.9/contextlib.py\", line 79, in inner
    return func(*args, **kwds)
  File \"/contentcuration/contentcuration/contentcuration/utils/publish.py\", line 842, in publish_channel
    kolibri_temp_db = create_content_database(channel, force, user_id, force_exercises, progress_tracker=progress_tracker)
  File \"/contentcuration/contentcuration/contentcuration/utils/publish.py\", line 130, in create_content_database
    tree_mapper.map_nodes()
  File \"/contentcuration/contentcuration/contentcuration/utils/publish.py\", line 200, in map_nodes
    self.recurse_nodes(self.root_node, {})
  File \"/contentcuration/contentcuration/contentcuration/utils/publish.py\", line 237, in recurse_nodes
    self.recurse_nodes(child, metadata)
  File \"/contentcuration/contentcuration/contentcuration/utils/publish.py\", line 237, in recurse_nodes
    self.recurse_nodes(child, metadata)
  File \"/contentcuration/contentcuration/contentcuration/utils/publish.py\", line 237, in recurse_nodes
    self.recurse_nodes(child, metadata)
  File \"/contentcuration/contentcuration/contentcuration/utils/publish.py\", line 229, in recurse_nodes
    exercise_data = process_assessment_metadata(node, kolibrinode)
  File \"/contentcuration/contentcuration/contentcuration/utils/publish.py\", line 482, in process_assessment_metadata
    exercise_data = extra_fields.get('options').get('completion_criteria').get('threshold')
AttributeError: 'NoneType' object has no attribute 'get'

Expected behavior

1. We get alerted to these types of errors
2. We have better logging and defensive checks against malformed data (which get is the problem and on which node?)

[bjester]

Followup issue added and PR for immediate fix merged #4174