Optimize checking internal link hashes in target files

[riccardoporreca]

By collecting non-duplicate hashes by file and calling the time-consuming create_nokogiri() only once per file.

Should address #768

See the attached output (htmlproofer-output.txt) of running the internal link checks on the build-site site artifact discussed in #768 (comment), via

htmlproofer --disable-external true --log-level debug --checks Links  build-site/

• The total execution is now around 10 min
• The output shows also the additional info / debug logs that help tracking how much htmlproofer is doing for checking internal links / hashes
  • in particular, it reflects the new approach of checking internal hashes by target file
  • more in general, it reports at debug level how many individual references to the same internal URL ("metadata") we iterate we have to iterate over

[gjtorikian]

Wow. Beautiful. Thank you so very much for this work~!

[riccardoporreca]

My pleasure @gjtorikian!