Enhance handling of invalid UTF-8 sequences in check_numeric_host function

[coderabbitai]

Following the discussion in PR #1, there's a consensus on improving the function to gracefully handle invalid UTF-8 sequences. The '�' character used to replace invalid sequences should potentially be filtered out.

[sporkmonger]

As mentioned in #1, the from_utf8_lossy function should help filter out invalid unicode sequences in a way that's difficult to use as a detection bypass.