Fixed 2 games both with empty collections resulting in 0 similarity

[Jeshibu]

This fixes #469

I have verified that:

• [x] These changes work, by building the extension and testing.
• [x] That the changes comply with the rules indicated in the repository.
• [x] Pull request is targeting master branch.

[darklinkpower]

Changes like this should be done in the control game matching algorithm and not to that method, since it calculates the Jaccard Similarity and this change would make it inaccurate. I'm also not sure if the best approach is to give a fixed number in such cases but I'm open to suggestions to improve the algorithm and if you've checked that it works fine.

[Jeshibu]

I've changed it so that it only uses fields that have values for at least one of the games. Fields with 0 (non-ignored) items in both lists now do not contribute to the similarity score, but also don't contribute to the minimum similarity score threshold.

I've also tweaked the minimum similarity threshold, but I want to put that in an advanced settings tab, along with the field weights, if you think this is the right way to go.

[Jeshibu]

Added settings. If this is about right I'll add localization too.