Enforce reset in each OpQueue subtable

[ledwards2225]

Each ECC op subtable concatenated in the merge protocol must be constrained to start (or end) with a reset operation so that the accumulator value from one table does not effect the value of the next.

One idea was to do this check automatically in the ECCVM, triggered by a zero row. It could also be added manually as part of the main protocol and enforced in the ecc op relation.

[arielgabizon]

I think it's enough our kernel circuits start with reset. If some app wants to use ec ops in a weird/incorrect way and not reset, that's their choice. It doesn't affect us/other apps if we/they make sure to start with a reset. So I don't think this needs to be enforced in the ECCVM. That would be pretty hard to do anyways, if the lengths of different subtables are not known in advance.

[ledwards2225]

In light of Ariel's observation (which makes sense), we should simply make sure kernels have this property.

[ledwards2225]

We've agreed that enforcing an eq_and_reset at both the beginning and end of all kernel subtables is sufficient. If someone writes an app that does not terminate in a reset, this could lead to failure in the ECCVM since the initial eq_and_reset in the subsequent kernel could fail to be satisfied - this falls into the category of "apps can be junk as long as they can't help forge proofs for other circuits".

[ledwards2225]

Closed with AztecProtocol/aztec-packages#15734