Make it so expand filter can take similar-limit as a second part

[jkomoros]

Originally captured in #570.

For the second part of the expand filter, it currently only takes children/descendants/etc filters. But it should also make it possible to return cards close to the given set that are above some similiarity threshold.

similar doesn't actually filter anything, so we'd need to add a similar-limit filter that takes a card-id as well as a similarity cutoff float. And then when it's used as the second part of the expand filter, it, like the children filters, ignores the card-ID and passes in the full set of cards from the first part

[jkomoros]

The similar filter should be able to take a combination of all of the cards enumerated, and then combine their fingerprint (kind of like a tag) and then find things that are similar to that fingerprint.

[jkomoros]

• [x] Add a similar-cutoff filter that takes a list of cards and a float cutoff
• [x] Support URL_PART_FLOAT
• [x] parseKeyCardID should return cardID, includeKeyCard, allCardIDs, so it can handle multiple key cards.
• [x] keyCardID should, if the cardID is an array, join it with union
• [x] Make it so similarCutoff can take slugs not just IDs
• [x] Use parseKeyCardID in similar, and similarCutoff constructors
• [ ] Fix the problem where if you have a float argument in the URL and hard refresh the server breaks due to the '.' in the URL. (Is that just polymer serve or would that be the other ones too? if so then 🤷 ). It's only in polymer serve, and it can be skipped by just having the demo float be an int with no period
• [x] Make it so makeSimilarConfigurableFilter et al, if rawCardID contains multiple IDs, it handles them fine
• [x] Make it so similar-cutoff's filter generator can also take an override list of card ids (just like the makeCardLinks can)
• [x] Make it so expand can accept either a card links thing or a similar-cutoff thing. (Redefine URL_PART_LINKS_FILTER)
• [ ] Consider making it so expand also calls the card BFS in the same way we do for similar card, to factor out more. (if we do, remove the whole "you can pass an override list of IDs" for BFSMaker and just unpack the cardIDList string). This is actually a pain to do mainly because there are lots of different filters each with a different set of arguments and different order, and figuring out how to pack them in, vs going straight to cardBFS, is annoying.

[jkomoros]

This is now mostly done except for a couple of niggling cleanup issues.