Host-link extraction: preserve `www.` prefix

[sebastian-nagel]

The CCF host-level web graphs (since 2017) were created with the leading www. stripped from the host name. Unlike the SURT-normalization, the prefix is only stripped

• if at least two dot-separated segments are preserved (www.com is kept intact)
• no www1. prefixes are stripped.

The reason for the stripping is the reduced size of the host-level web graphs:

• Back in 2017 this made the graphs approx. 10% smaller.
• This number went down over the years and the storage saving are now only about 5%.

While storage benefit became smaller, the stripping has a couple of disadvantages:

1. Joining web graph data with other host-level data is more difficult.
2. Fetching the homepage of a stripped host name requires to follow a redirect, or may even fail or return a different result.
3. Extra documentation is required, in addition to the reverse domain name notation.

Starting with the first web graph in 2026 (cc-main-2025-26-nov-dec-jan-host), the leading www. in a host name will be preserved.