* Separate types for Relaxed R1CS and R1CS instances and witnesses
* Allows creating default values for Relaxed R1CS types
* StepSNARK now folds a regular R1CS instance-witness into a running Relaxed R1CS instance-witness
* We additionally enforce input chaining checks: the incoming instance must have input that matches the output of the incremental computation thus far

* Support pasta curves directly.

* Upgrade rand.

* Simplify.

* Don't deny missing docs.

Co-authored-by: porcuquine <porcuquine@users.noreply.github.com>

and cleanup some group op traits