EntryTemplate.extract_search_constraints now returns constraints for
"(|(FILTER))" as if it were "(FILTER)".

An AttributeError was raised in match_equal/match_substr of string matching
rules if any attribute value contained a prohibited character. These attribute
values are now ignored for matching.

Minimal supported python version is changed to 3.7 to support Debian Buster.

DN/RDN/RDNAssertion now require a schema parameter that is used for correct
parsing and comparing. DN and RDN objects are now longer hashable (this may
change later). Hex-formatted assertion values in DN strings are still not
supported.

Attribute type and object class definitions are now parsed from their string
representation. Extensible filters are now supported.

Somewhat unrelated: Object.match_search and Object.get_search_result_entry
are replaced with Object.search. Object.compare_dn and Object.match_compare
are replaced with Object.compare.

Schema and object code now has extensive unit tests.

This control extension is enabled by default and works with existing
do_search handlers .

The default implementation relies on do_search to get the requested object.

Matching rules objectIdentifierMatch and objectIdentifierFirstComponentMatch
now behave correctly when comparing a numeric OID and the corresponding short
descriptive name or two different short descriptive name belonging to the
same OID.

Full schema support

See merge request !2

DN handling now fully conforms to RFC4514 and has a high test coverage.
Still lacks a lot of documentation.