(MT) data cleanup analysis

For both the recommender system ("recys") as well as for the MT, we need to operate on clean data. Otherwise, it's garbage in->garbage out.

• take a sample of 100 ContentItems.{content, title, summary} fields. Look at them. Identify weird data (example: &#8282 repeated strings)
• Put them into categories of artefacts
• Decide which ones to keep and which ones to clean up
• Think again about what to clean up with respect to the recsys
• Implement it as a python library
• run it on a DEV repco instance