Fix lexing of string literals ending in an escaped backslash

[gaoflow]

The lexer's quoted-string patterns locate the closing quote with a single-character look-behind:

self.double_quote_pattern = r'"(?P<G_DQUOTE>(?:(?!(?<!\\)").)*)"'
self.single_quote_pattern = r"'(?P<G_SQUOTE>(?:(?!(?<!\\)').)*)'"

(?<!\\) cannot tell \\" (an escaped backslash followed by the real closing quote) from \" (an escaped quote), so for a name selector whose key ends in an escaped backslash the pattern runs past the real closing quote and a following selector then mis-pairs with a later quote.

The visible effect is that JSONPath produces a normalized path it cannot itself read back, which violates RFC 9535 §2.7 (a normalized path must select exactly its node):

>>> import jsonpath
>>> jsonpath.findall(r"$['a\\']['b']", {"a\\": {"b": 1}})
jsonpath.exceptions.JSONPathSyntaxError: unexpected token "'"

That $['a\\']['b'] is python-jsonpath's own generated normalized path for the node, so findall(match.path, doc) round-trips break.

The fix replaces the look-behind with the standard disjoint "unescaped char or backslash-escape" form, which is unambiguous and backtracking-free:

self.double_quote_pattern = r'"(?P<G_DQUOTE>(?:[^"\\]|\\.)*)"'
self.single_quote_pattern = r"'(?P<G_SQUOTE>(?:[^'\\]|\\.)*)'"

This is the same look-behind bug class fixed in #124 for the regex-literal pattern; that change left the two string-literal patterns untouched. The re_pattern regex-literal pattern is out of scope here.

tests/test_escaped_backslash_string_literals.py covers lexing/parsing of keys ending in escaped backslashes plus normalized-path round-trip and idempotence over several documents. Full suite passes (2793), ruff and mypy --strict clean.

[jg-rp]

Thanks @gaoflow 👍