Filter-rewriting 2: rewrite harder

[Zac-HD]

#2701 proposed "rewriting" filters in simple cases, i.e. modifying the strategies to satisfy the condition by construction rather than rejection sampling. We shipped this for numeric comparisons, as well as some string methods and float properties like isnan, and it's been working well!

Following #3780, we'd like a way to do filter-rewriting for length filters.

• starting with partial(min_len, x) for the min_len and max_len functions in filtering.py
• as an extension, using AST analysis to cover e.g. lambda ls: 0 < len(ls) <= 10 - in this case we'll want to apply the rewrite for efficiency, and apply the predicate as a filter in case the name len has been rebound in the enclosing scope.
• support st.lists(...).map(some_collection_type), in order to handle st.sets(), st.binary(), st.dictionaries(), and similar user-defined strategies (including .map(tuple), which is returned by st.from_type(tuple[T, ...])).
  • it's probably better to handle this in ListStrategy.map than in MappedStrategy.filter? Gonna be an inelegant hack either way, but the results should be worth it.

There have also been a variety of further filtering extensions proposed in older issues, which I'm listing here in case someone would like to implement one of them. Note however that they're lower priority, so unless someone has explicitly volunteered I'll close this issue when the length filters are done.

• Comparing elements of tuples() or fixed_dictionaries() to constants, like lambda d: 1 <= d[0] <= 10, by pushing a filter down onto the specific element. If we use the partial(operator.x, value) syntax then the elements strategy will also rewrite itself too!
• Following Enhanced st.from_regex() strategy with alphabet=... argument and filter-rewriting #3479 (comment), st.text()/st.binary() filtered with re.compile(...).find/match/fullmatch
  • st.from_regex() with the same filters - note that this will require some upstream work in greenery; see also Use greenery and regex_transformer to merge pattern and patternProperties keywords python-jsonschema/hypothesis-jsonschema#85
• Converting operator.contains / x.__contains__ lambda x: x in (1, 2, 3) into a st.sampled_from() strategy, for simple types like integers and strings