Fix '#' inside string literals being lexed as a comment - #22

Fixes #21.

Problem

The comment token runs to end of line, so when the lexer sits at the interior of a string like "#fff", the comment token (5+ chars, through the closing quote and beyond) outmatches the string-content token (4 chars, stops at the quote) under the longest-match rule. The string interior is consumed as a comment and the parse of everything after it is corrupted — with a later string in the file, the quotes re-pair across rules:

package p

color := "#fff"

allow := true

parses today with #fff" as a (comment) inside an ERROR.

Fix

Give the quoted_string and raw_string content tokens lexical precedence (prec(1)) over the comment token — exactly the precedence the two interpolated string forms (interpolated_string_double, interpolated_string_raw) already carry, which is why $"#..." was unaffected. Lexical precedence takes priority over match length, so string content wins inside string interiors, while comments after strings on the same line still lex normally (string-content tokens aren't valid outside string interiors):

x := "#a" # real comment

keeps (quoted_string) followed by (comment).

src/ is regenerated with tree-sitter generate.

Tests

New test/corpus/strings.txt covering # in quoted strings, # in raw strings, and a real comment following a string that contains #. Full corpus: 42/42 passing.