fix: normalize direct fetcher URLs

[chaliy]

What

Normalize bare domain URLs before parsing inside every built-in public fetcher implementation, so direct Fetcher::fetch calls behave like the registry/tool/client paths.

Why

The previous bare-host fix covered normal fetchkit execution paths, but public fetcher structs could still be called directly with a bare URL and fail at Url::parse(&request.url).

How

• Added a normalized-copy helper for borrowed FetchRequest values.
• Normalize at the top of every built-in fetcher fetch() implementation and in DefaultFetcher::fetch_to_file().
• Added direct-fetcher regression tests using the in-memory transport.
• Updated stale invalid-URL wording and specs to reflect accepted bare domain URLs.

Risk

• Low
• Direct fetcher calls that previously returned InvalidUrlScheme for domain-like bare URLs may now proceed to fetch over HTTPS.

Checklist

• Unit tests are passed
• Smoke tests are passed
• Documentation is updated
• Specs are up to date and not in conflict
• Formatting, clippy, tests, docs, and release build smoke passed locally