introduce fts_term() function for preparing tokens for FTS5 token matching

[missinglink]

I noticed what appears to be a bug in the code but appears to have no negative effects on the quality of results.

There is a piece of code which replaces all occurrences of space with underscore (.replace(/ /g, '_')).

I don't remember exactly why we do this but I assume its to allow us to keep multi-word tokens (such as 'new york') as a single token.

Recently we started using the FTS index more, so we should apply this space->underscore replacement in all parts of the code which use the FTS index.

Now.. why it 'just works' anyway is a bit of a mystery 🤷

• https://github.com/pelias/placeholder/blob/master/lib/TokenIndex.js#L25-L34
• https://github.com/pelias/placeholder/blob/master/lib/TokenIndex.js#L54

[orangejulius]

I just rebased this after fixing the tests in #221, do we still want to merge it? I don't recall if we tested the changes in this PR or what we think it might have helped with. Performance maybe?