Add fuzzy matching during submission processing to correct spelling mistakes

[spwoodcock]

Is your feature request related to a problem? Please describe.

• Sometimes the user types in the name of something in ODK Collect slightly wrong.
• For example in Bali we mapped buildings inside of compounds. When providing the compound names as a free text entry, users entered similar names that only varied slightly.

Describe the solution you'd like

• We could probably use a library like thefuzz (already in osm-fieldwork I believe) for fuzzy matching of strings.
• We could lump together similar string values, which would provide better insights into data when it is nicely processed/cleaned.

Describe alternatives you've considered

• Of course, there is a risk of false positives and incorrectly modifying values.
• We should set a high threshold for matching, so only very similar words can match.