Consider adding a Unicode normalization storage transformer

[jstriebel]

Since unicode characters are allowed for node names unicode normalization is a useful feature, which is also recommended for users. It would be great to add a group storage transformer which normalizes all node-names of sub-groups and sub-arrays, so that only normalized names are stored and they can be retrieved by using any string that normalizes to the same form.

As recommended by the Unicode Standard Annex # 31 we also recommend to use the normalization form KC (NFKC) in the spec, which should also be used by this transformer (at least by default, there may be reasons to allow customizing this). NFC might also be used to fit filsystem storage normalization, see #56 (comment).

The main discussion around this topic so far happened in issue #56.