CESR followups

[squell]

• Attachment groups and index groups missing from signatures (see Section 9.3)
• Count codes: currently a dummy value is provided, please see if an accurate quadlet/triplet count can be substituted instead
• Remove all nonconfidential_data from encrypted-and-signed envelope

• We also strongly suggest that in the TSP implementors draft, section 7.1.5, the signature is not over the "whole preceding message", for two reasons:

• this is "late", we are already busy encoding the message and then we have to produce a signature in the middle of that, it's better to write the spec so that all of these things can be done 'discretely'
• For a spec, "the preceding message" is a rather imprecise stipulation and runs the risk of being interpreted very differently by different implementors

Luckily the fix is very easy: make the signature over the Digest which is also embedded in that message. That way the signature can be made before starting encoding the message, and it is also unambiguous. And it achieves the same point: proving that the VID has control over the signing key.

• We also strongly suggest to not use XRFI (TSP_RFI) / XRFA (TSP_RFA) for several control messages, so a TSP message decoder can more easily distinguish between the type of control message (direct request, parallel request, nested request...). This will also allow us to simply tick off the last bullet item above. (The current SDK uses XRNI and XRNA for nested relationship control messages, but does use XRFI/XRNA for both direct and parallel releationship forming--but I would consider this rather brittle)