Different possible optimizations for count_q and count_v

[rikard-sics]

We had a number of proposals about optimizations for the counters.

• Use the device clock in a clever way, instead of saving counters
  • Proposal by Carsten during IETF 112 and interim 2022-04-28
• Constrained devices saving count_q/count_v regularly
  • Similar to OSCORE Appendix B.1
• Each client can count q up to half the q limit
  • Proposal from Christian
  • See notes here: https://hackmd.io/sOliRiyGQT-OmjxbprN9XA
  • What is the maximum overestimation?

We should consider how these 3 proposals work, and how they would interact with each other or can be combined. We can also clarify that saving the counters every N is just a suggestion for implementers. Different solutions may be acceptable. The key point is making sure to not lose or underestimate the counters after reboot (while overestimating them can be fine to some extent).