Potentially wrong constraints in `GlReduce`

[shuklaayush]

Shouldn't the Reduce function constrain that both r and d are less than the Order() p instead of just constraining that they're less than 2^64?

plonky2-circom/circom/circuits/goldilocks.circom

Lines 21 to 37 in 806f6a4

	// Gl: Goldilocks
	// range check d < 1 << N
	template GlReduce(N) {
	signal input x;
	signal output out;
	var r = x % Order();
	var d = (x - r) \ Order();
	out <-- r;
	signal tmp0 <-- d;
	tmp0 * Order() + out === x;
	component c0 = LessNBits(N);
	c0.x <== tmp0;
	component c1 = LessNBits(64);
	c1.x <== out;
	}

Eg. If x is p + 1, then both (d, r) = (1, 1) and (0, p + 1) are valid witnesses with the current code