starting work on reverse matching

[nrktkt]

current state is just hacked together POC with some notes

need to integrate with damnhandy test cases

challenges

unreserved character encoding lives at the expression level, but we need that
info to build regex at varspec level

matching of fragment and unreserved characters in list is seemingly impossible
since , is allowed in fragments, but is also the separator token. The RFC seems
to note this so most likely the solution is to document that we don't support
values with reserved delimiters

does order need to be considered for key-value paired expressions like ; ? & ifso
named capture groups cannot be used here
idea:
expressions surface two methods, getMatchPattern() and Map getMatches(Matcher)
getMatchPattern returns a regex to match that expression, but rather than using
named capture groups which can be used at the top level, the expression uses
getMatches to return a map from variable names to values (which can internally use
ncg, or not).

ex: {/var,x}/here{?x,y,empty} might generate pattern like
(?<TWFuIGlzI>(?<var>\/[unreserved]*)(?<x>\/[unreserved]*))\/here(?<GRpc3Rp>\?((x|y|empty)=([unreserved]*)&?){0,3})
the expression {/var,x} would grab the group TWFuIGlzI and use the known group
names with ncg to get var and x, while {?x,y,empty} would look at each key value group tuple
#2