Replace StringScanner

[andyundso]

This PR replaces the StringScanner in unescape with regular gsub. In order for it to not match too much, I needed to add a negative regex lookahead. Performance is still a bit better than before.

require 'benchmark/ips'
require_relative 'lib/rdf'

Benchmark.ips do |x|
  x.report('without') do
    RDF::NTriples::Reader.unescape("D\u00FCrst" )
  end

  if ENV['WITH_MODULE'] == 'true'
    module RDF::NTriples
      class Reader
        UCHAR4          = /(?<!\\)\\(?!\\)u([0-9A-Fa-f]{4,4})/.freeze
        UCHAR8          = /(?<!\\)\\(?!\\)U([0-9A-Fa-f]{8,8})/.freeze

        def self.unescape(string)
          # Note: avoiding copying the input string when no escaping is needed
          # greatly reduces the number of allocations and the processing time.
          string = string.dup.force_encoding(Encoding::UTF_8) unless string.encoding == Encoding::UTF_8

          string
            .gsub(ESCAPE_CHARS_ESCAPED_REGEXP, ESCAPE_CHARS_ESCAPED)
            .gsub(UCHAR) do
              [($1 || $2).hex].pack('U*')
            end
        end
      end
    end
  end

  x.report('with') do
    RDF::NTriples::Reader.unescape("D\u00FCrst" )
  end
  x.hold! 'temp_results'
  x.compare!
end

ruby 3.4.4 (2025-05-14 revision a38531fd3f) +PRISM [x86_64-linux]
Warming up --------------------------------------
                with   250.703k i/100ms
Calculating -------------------------------------
                with      2.530M (± 2.5%) i/s  (395.26 ns/i) -     12.786M in   5.056920s

Comparison:
                with:  2530003.9 i/s
             without:  1478262.7 i/s - 1.71x  slower

[coveralls]

coverage: 91.721% (-0.01%) from 91.731%
when pulling 7622a28 on andyundso:remove-string-scanner
into cae8390 on ruby-rdf:develop.

[gkellogg]

Thanks, this looks good. I'll plan a new release soon.