If a site has "/" as its entrypoint but the languages
have FQDNs or subdomains or a mixture between subdomain
the base without domain always gets found and the subdomains
are not considered.
Currently this situation is handled in
SiteMatcher->getRouteCollectionForAllSites()
which does not fit a certain scheme if there is a mixture.
This change corrects this and adds tests to handle any cases:
- just / as base, and then /de and /en as language base
- just / as base, and then subdmoains or full domains+path as language base
- a full URL as base, and then path parts as language base
- a full URL as base and then full URLs + subdomains as language base
- Also check for http and http variants.