Significantly improve event matching/filtering

[ahobsonsayers]

Significantly improve event filtering/matching by improving the normalisation of the event names and by implementing a home rolled string similarity algorithm for event names.

The string similarity algorithm works by using a modified Smith-Waterman local alignment algorithm to align a substring (desired event name) with a target string (the actual even name), and then a optimal string alignment Damerau-Levenshtein to calculate similarity on a word level within the substring. Word similarity is then averaged across all the words, after best alignment, to give the final similarity score.

This home rolled algorithm seems very robust and extremely suitable for our use case. This can be seen in the tests for the algorithm.