Handle arbitrary nested NOT/AND/OR in queries

[ml-evs]

This PR allows us to recursively parse nested queries, but does NOT fix the problem described in #79 (yet...). Essentially the expression_phrase function fails to go deep enough through the tree and ends up flattening out the query at the first AND/OR, which breaks any negations.

An example of the tests failing against master can be found here, which amounts to NOT (expr1 AND expr2) becoming NOT expr1 and expr2 instead of NOT expr1 AND NOT expr2. This PR fixes that by extending the negation to the second expression in this example, and hopefully to arbitrary expressions.

The discussion in #79 amounts to the distinction between the above and NOT (expr1 AND expr2) being tricky to handle in Mongo as it stands, which instead would need something like (expr1 AND NOT expr2) OR (NOT expr1 AND expr2), which should probably be handle with care in another PR.

[codecov]

Codecov Report

Merging #221 into master will increase coverage by 0.02%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #221      +/-   ##
==========================================
+ Coverage   87.55%   87.57%   +0.02%     
==========================================
  Files          43       43              
  Lines        1912     1916       +4     
==========================================
+ Hits         1674     1678       +4     
  Misses        238      238

Flag	Coverage Δ
#unittests	87.57% <100%> (+0.02%)	⬆️

Impacted Files	Coverage Δ
optimade/filtertransformers/mongo.py	95.65% <100%> (+0.19%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2ed7f74...ee1a400. Read the comment docs.

[CasperWA]

Seems legit. Thanks.
You can leave out my comment changes if you want - especially the longer one. I just thought to make the negation more clear in the comment, but I see now that then the nor comment should change as well...

Furthermore, a couple of questions.

[CasperWA]

Seems good to me 👍

It's a shame about the quirks with Lark not permitting recursive methods on the ones it called. One could consider making the recursive method a function in the expression_phrase method ... but in the end its all the same, and this, honestly, looks prettier than my suggestion.

[ml-evs]

Seems good to me +1

It's a shame about the quirks with Lark not permitting recursive methods on the ones it called. One could consider making the recursive method a function in the expression_phrase method ... but in the end its all the same, and this, honestly, looks prettier than my suggestion.

Yeah, I considered having some kind of closure inside expression_phrase but I'd be more worried about unintended side effects if we did it that way.