make TermNormIsSubStringMappingStrategy handle multi-word substrings

[EFord36]

Closes #274

Currently, this will only prefer a term if the term norm is a single word contained (as a word) in the ent_match_norm. Change to also prefer it if the term norm is a sequence of words that are a substring of the ent_match_norm.

In draft because I still need to:

• Check impact on performance (we're using a regex match now, is this too slow? We could compare to looking for the first word with list.index, and then iterating. The implementation here is longer though. We could also try mypyc on that for interest).
• Check impact on behaviour (does this change anything in the test documents we have for the different use cases? It should help, but does it?)
• Write tests

That said, the actual code is ready to look at to assess 'is this a good idea' in a broad sense?