Improve extraction of host names and registered domains

[sebastian-nagel]

• no host name is extracted in the following situations
  • URL contains 4 slashes after the protocol: https:////example.org/ - while java.net.URL extracts an empty hostname, the Nutch's OkHTTP-based protocol seems to fetch the resource as if there are only two slashes.
  • similarly java.net.URL and OkHttp show a different behavior if there is an overlong (or even invalid?) userinfo before the hostname (scheme://userinfo@hostname/)
• IP addresses are not recognized as such if ending in a dot: https://123.123.123.123./robots.txt
• the extraction of registered domains (done by crawler-commons' EffectiveTldFinder does not extract anything if the hostname is equal to a public suffix (gov.uk, kharkov.ua for example)

[sebastian-nagel]

IP addresses are not recognized as such if ending in a dot: https://123.123.123.123./robots.txt

This has been addressed in #25. Added unit test in d5a7f74.

Extraction of registered domains from host names equal to a registry suffix persists (both gov.uk and kharkov.ua):

select count(*) as cnt,
       url_host_name,
       url_host_registered_domain
from ccindex
where crawl = 'CC-MAIN-2026-04'
  and contains(array['gov.uk','kharkov.ua'], url_host_name)
group by url_host_name, url_host_registered_domain;