Tinygrad - not supported op

[janchk]

While trying to run tinygrad's OnnxRunner (main branch) stumbled across op='NonZero' is not supported error in vision part of smolVLA module.
This module -> https://huggingface.co/ainekko/smolvla_libero_onnx/blob/main/smolvlm_vision.onnx
Need to locate this op in graph and decide what to do next.
Either extend Tinygrad or do some hack over existing onnx graph to remove this op.

FYI
List of supported onnx ops in tinygrad

['Neg', 'Reciprocal', 'Pow', 'Sqrt', 'Sign', 'Abs', 'Exp', 'Log', 'Mish', 'Sin', 'Cos', 'Tan', 'Asin', 'Acos', 'Atan', 'Relu', 'Sigmoid', 'MatMul', 'Floor', 'Ceil', 'IsNaN', 'Softplus', 'HardSwish', 'Where', 'Mul', 'Sinh', 'Cosh', 'Tanh', 'Softsign', 'Asinh', 'Acosh', 'Atanh', 'Elu', 'Celu', 'Selu', 'Round', 'Erf', 'If', 'Identity', 'Constant', 'Range', 'ImageDecoder', 'EyeLike', 'OptionalHasElement', 'OptionalGetElement', 'ConstantOfShape', 'Size', 'Shape', 'Not', 'Clip', 'IsInf', 'HardSigmoid', 'BiasGelu', 'FastGelu', 'PRelu', 'LeakyRelu', 'ThresholdedRelu', 'LogSoftmax', 'Binarizer', 'Add', 'Sub', 'Div', 'Less', 'LessOrEqual', 'Greater', 'GreaterOrEqual', 'Equal', 'And', 'Or', 'Xor', 'BitwiseAnd', 'BitwiseOr', 'BitwiseXor', 'BitwiseNot', 'Mod', 'Cast', 'CastLike', 'Max', 'Min', 'Mean', 'ReduceMax', 'ReduceMin', 'ReduceMean', 'ReduceSumSquare', 'ReduceProd', 'ReduceL1', 'ReduceL2', 'ReduceLogSum', 'ReduceLogSumExp', 'ArgMin', 'Reshape', 'Flatten', 'Expand', 'Shrink', 'Transpose', 'Squeeze', 'Unsqueeze', 'Tile', 'Concat', 'Slice', 'Split', 'Pad', 'CenterCropPad', 'AveragePool', 'MaxPool', 'ConvTranspose', 'MaxUnpool', 'GlobalMaxPool', 'Gemm', 'Einsum', 'CumSum', 'Trilu', 'Upsample', 'TopK', 'BatchNormalization', 'InstanceNormalization', 'LayerNormalization', 'SkipLayerNormalization', 'EmbedLayerNormalization', 'MeanVarianceNormalization', 'OneHot', 'DepthToSpace', 'SpaceToDepth', 'LRN', 'NegativeLogLikelihoodLoss', 'SoftmaxCrossEntropyLoss', 'AffineGrid', 'RMSNormalization', 'RotaryEmbedding', 'ArrayFeatureExtractor', 'Gather', 'Scatter', 'GatherND', 'ScatterND', 'GatherElements', 'Compress', 'QuantizeLinear', 'DynamicQuantizeLinear', 'DequantizeLinear', 'QLinearConv', 'QLinearMatMul', 'QLinearAdd', 'QLinearMul', 'QLinearGlobalAveragePool', 'ConvInteger', 'MatMulInteger', 'Adagrad', 'Adam', 'Momentum', 'Gradient', 'ArgMax', 'Conv', 'Gelu', 'GlobalAveragePool', 'GroupNormalization', 'ReduceSum', 'Resize', 'ScatterElements', 'Sum', 'Softmax', 'Dropout', 'Attention']