[test_pattern_match_edgecases.jl] Add stronger HardFail inference for pattern_match on 3 or more VarNodes

[Whebon]

Variable :x is assigned to the first RuleNode r1 encountered.
Other matches with variable :x are then pairwise compared to RuleNode r1.
Any SoftFails involved are not combined to a HardFail.
The tests below demonstrate HardFails that are not caught by the pattern_match:

    @testset "3 VarNodes: pairwise Softfail, triplewise HardFail" begin
        #TODO: this test fails because, in the current implementation, variable comparisons are done pairwise
        # domains of holes within vars should be updated for stronger inference
        rn = RuleNode(4, [
            RuleNode(4, [Hole(BitVector((1, 1, 0))), Hole(BitVector((0, 1, 1)))]), 
            Hole(BitVector((1, 0, 1)))
        ])
        mn = RuleNode(4, [
            RuleNode(4, [VarNode(:x), VarNode(:x)]),
            VarNode(:x)
        ])
        @test pattern_match(rn, mn) isa HerbConstraints.PatternMatchHardFail
    end

    @testset "3 VarNodes: HardFail on instance 2 and 3" begin
        #TODO: this test fails because, in the current implementation, only (1, 2) and (1, 3) are compared
        # domains of holes within vars should be updated for stronger inference
        rn = RuleNode(4, [
            RuleNode(4, [Hole(BitVector((1, 1, 1))), RuleNode(1)]), 
            RuleNode(2)
        ])
        mn = RuleNode(4, [
            RuleNode(4, [VarNode(:x), VarNode(:x)]),
            VarNode(:x)
        ])
        @test pattern_match(rn, mn) isa HerbConstraints.PatternMatchHardFail
    end

Why is this not here yet

Forbidden constraints with 3 or more occurrences of the same VarNode are considered an edgecase.