Path#relative_to? across equivalent Windows roots

[HertzDevil]

Path#relative_to? requires both the target and the base paths to have the exactly same anchor, otherwise nil is returned. It might be surprising to some that alternative Windows directory separators are not considered:

Path["C:/foo"].relative_to?(Path["C:\\foo"]) # => nil

Resolving this part is relatively easy; as long as the drive letters are identical, it should not matter which slash is used.

The complicated part arises when dealing with local device paths (#15590), because either the target or base can start with \\., and it could be equivalent to some other normal or UNC path:

Path["//./C:/foo"].relative_to?(Path["C:/bar"]) # => "../foo"
Path["//./D:/foo"].relative_to?(Path["C:/bar"]) # => nil

Path["//./UNC/server/share/foo"].relative_to?(Path["//server/share/bar"]) # => "../foo"

Path["C:/foo"].relative_to?(Path["//./C:/bar"]) # => "../foo"
Path["D:/foo"].relative_to?(Path["//./C:/bar"]) # => "../../D:/foo"

Path["//server/share/foo"].relative_to?(Path["//./C:/bar"]) # => "../../UNC/server/share/foo"

These can be done purely lexically, without any filesystem access. So I wonder if Path#relative_to? should take care of these cases as well.