bug: Solver consistency test fails for commitment circuit with small modulus

[ivokub]

Description

In test TestSolverConsistency for circuit commitCircuit we get test failures when checking commitment != 0. But when we use the small field (modulus 47), then the probability of commitment being randomly 0 is high and we hit it with high probability (and in practice as inputs are fixed).

Expected Behavior

TestSolverConsistency/commit succeeds

Actual Behavior

It doesnt:

errSCS :<nil>
        errR1CS :<nil>
        errEngine(const=false): [assertIsDifferent] 0 != 0
        test.isSolvedEngine.func1
        	gnark/test/solver_test.go:204
        runtime.gopanic
        	go1.21.0/src/runtime/panic.go:914
        test.(*engine).AssertIsDifferent
        	gnark/test/engine.go:450
        circuits.(*commitCircuit).Define
        	gnark/internal/backend/circuits/commit.go:16
        
        errEngine(const=true): [assertIsDifferent] 0 != 0
        test.isSolvedEngine.func1
        	gnark/test/solver_test.go:204
        runtime.gopanic
        	go1.21.0/src/runtime/panic.go:914
        test.(*engine).AssertIsDifferent
        	gnark/test/engine.go:450
        circuits.(*commitCircuit).Define
        	gnark/internal/backend/circuits/commit.go:16
        
        witness: [1, 2]

Possible Fix

We could build the commitment in test engine such that we avoid 0 outputs, but that would be cheating. I would rather omit running consistency test for this circuit. But the consistency test uses all integration tests, so we need to have some option to avoid running some of the tests for some circuits. This is also useful for some other edge cases (imo there was also some issue with some other circuits where we have hardcoded which ones do avoid).

[ivokub]

Fixed in #919