Extend HostToDomainGraph to fold host-level graphs stripping the `www.` prefix

[sebastian-nagel]

Per commoncrawl/cc-pyspark#56 the Common Crawl web graphs preserve a www. prefix in the host name. A tool to convert the graph including the www. prefix to one without it, would be useful to compare 1:1 how this changed the graph structure and derived properties, such as the centrality rankings.

The class HostToDomainGraph already supports two aggregation levels:

• (by default) registered domain - the domain name below the ICANN registry suffix defined by the public suffix list (PSL)
• "private" domain (command-line flag --private-domains) - the domain name below any public suffix including those in the private section of the PSL.

Adding a third aggregation level "host without www.", is simple:

1. Implement the stripping of the www. prefix following how it was done in cc-pyspark (see Host-level link extraction: preserve the www. prefix in host names cc-pyspark#57).
2. Expose this aggregation level per command-line options.
   • Because two mutually exclusive boolean flags are cumbersome, we might refactor the code to use the option --aggregation-level with three supported values: registered-domain (default), private-domain, host-without-www.
   • But ensure backward-compatibility of the current options, to avoid that scripts and documentation need to be adapted immediately.
3. Update the Javadoc and command-line help accordingly. Add note that the new aggregation level "stretches" the definition of a "domain".

[sebastian-nagel]

Implemented in #30. Thanks, @lfoppiano!