Multiple language versions for the same attribute

Closed with documentation added.

http://svn.shibboleth.net/view/cpp-sp?rev=3604&view=rev

Fixed schema, allowed for metadata attribute lang awareness by option to disable decoder caching.

http://svn.shibboleth.net/view/cpp-sp?rev=3603&view=rev

Lang awareness implemented in new resolver, extractor, and decoder APIs and decoders all support langAware option to select only the best matching value.

I think the better idea is to implement language matching for the values so that the best matching value is used to populate a single valued header. I can factor the logic added for metadata content extraction of language-specific information out into a reusable layer and then add lang awareness to the decoders.

The use case of extracting more than one value and tracking the language could be met by using the non-string decoder options in the SP to encode the lang attribute into the header value. Using different header names would be impossible, since the names would have to depend on the language tag, and it would be trivial to spoof since the set of tags is approximately infinite.

Per my list comment, I don't know what this really means in terms of software requirements. I'd suggest opening a thread on that on the dev list.