Add level 1 optimized mnist model so that the required_ops.config includes the ops in that (which are used in mnist.level1_opt.ort). NNAPI unit tests need this. (#6514)

Update required_ops.config with automatically generated version. There have been some other changes to testdata which show up as diffs here.

onnxruntime/test/testdata/mnist.readme.txt

@@ -0,0 +1,6 @@
+The mnist model is used in a multiple tests for minimal/mobile builds in both ONNX and ORT formats.
+
+We also save both ONNX and ORT format versions of the model with level 1 (aka 'basic') optimizations applied.
+  - mnist.level1_opt.onnx makes sure the required operators for this model are automatically included in 
+    required_ops.config, which is used in the reduced ops CI build.
+  - mnist.level1_opt.ort is used in NNAPI unit tests.

onnxruntime/test/testdata/required_ops.config

@@ -1,10 +1,10 @@
 # required ops for ONNX models in testdata
-ai.onnx;7;Abs,Add,And,BatchNormalization,Concat,Conv,Dropout,Flatten,Foo,Gather,Gemm,GlobalAveragePool,Identity,LSTM,MatMul,Max,MaxPool,Min,Mul,Mul16,Multinomial,OptionalOp,RNN,Relu,Reshape,Shape,Slice,Softmax,Sub,Unsqueeze
-ai.onnx;8;Add,Conv,Flatten,MatMul,MaxPool,Mul,Relu,Reshape
+ai.onnx;7;Abs,Add,And,BatchNormalization,Concat,Conv,Dropout,Flatten,Foo,Gather,GlobalAveragePool,Identity,LSTM,MatMul,Max,MaxPool,Min,Mul,Mul16,Multinomial,OptionalOp,RNN,Relu,Reshape,Shape,Slice,Softmax,Sub,Unsqueeze
+ai.onnx;8;Add,Conv,Flatten,Gemm,MatMul,MaxPool,Mul,Relu,Reshape
 ai.onnx;9;Abs,Add,BatchNormalization,Cast,Clip,Concat,Constant,ConstantOfShape,Conv,Div,Equal,Gather,Gemm,Identity,If,LayerNormalization,LeakyRelu,MatMul,Mul,Pow,ReduceMean,Relu,Reshape,Scan,Shape,Sigmoid,Slice,Softmax,Softsign,Sqrt,Sub,Tanh,Transpose,Unsqueeze
-ai.onnx;10;Add,Cast,Concat,Constant,ConstantOfShape,Div,Dropout,Equal,Erf,Expand,Gather,Gemm,Greater,Identity,If,LayerNormalization,Loop,MatMul,Min,Mul,Neg,NonZero,Not,Pow,ReduceMean,ReduceSum,Reshape,Shape,Slice,Softmax,SparseSoftmaxCrossEntropy,Sqrt,Squeeze,Sub,Tanh,TrainableDropout,Transpose,Unsqueeze
+ai.onnx;10;Add,Cast,Concat,ConstantOfShape,Div,Dropout,Erf,Expand,Gather,Greater,Identity,If,LayerNormalization,Loop,MatMul,Mul,Neg,NonZero,Pow,ReduceMean,ReduceSum,Shape,Sqrt,Squeeze,Sub,Tanh,Transpose,Unsqueeze
 ai.onnx;11;Abs,Add,ArgMax,BatchNormalization,Cast,Clip,Concat,Constant,ConstantOfShape,Conv,Div,Exp,Expand,Flatten,Gather,Gemm,Identity,If,LayerNormalization,Log,Loop,MatMul,MatMulInteger,Max,Min,Mul,Neg,Pow,RandomUniform,Range,ReduceMean,ReduceSum,ReduceSumSquare,Relu,Reshape,Scan,SequenceConstruct,SequenceInsert,SequenceLength,Shape,Sigmoid,Slice,Softmax,Split,Sqrt,Squeeze,Sub,Sum,Tanh,Transpose,Unsqueeze,Where
-ai.onnx;12;Add,And,Cast,Concat,Constant,ConstantOfShape,CumSum,Div,Dropout,DynamicQuantizeLinear,Equal,Erf,Expand,Flatten,Gather,GatherND,Gemm,GlobalAveragePool,Identity,If,LayerNormalization,Less,Loop,MatMul,MatMulInteger,Min,Mul,NonZero,Not,Pad,Pow,RandomNormalLike,Range,ReduceMean,ReduceSum,Reshape,Shape,Slice,Softmax,SoftmaxCrossEntropyLoss,Split,Sqrt,Squeeze,Sub,Tanh,TrainableDropout,Transpose,Unsqueeze,Where
+ai.onnx;12;Add,And,Cast,Concat,Constant,ConstantOfShape,Conv,CumSum,Div,Dropout,DynamicQuantizeLinear,Equal,Erf,Expand,Flatten,Gather,GatherND,Gemm,GlobalAveragePool,Identity,If,LayerNormalization,Less,Loop,MatMul,MatMulInteger,Min,Mul,NonZero,Not,Pad,Pow,RandomNormalLike,Range,ReduceMean,ReduceSum,Relu,Reshape,Shape,Slice,Softmax,SoftmaxCrossEntropyLoss,SparseSoftmaxCrossEntropy,Split,Sqrt,Squeeze,Sub,Tanh,Transpose,Unsqueeze,Where
 ai.onnx;314159;Add
 ai.onnx.ml;1;ArrayFeatureExtractor,DictVectorizer,LabelEncoder,LinearClassifier,Normalizer,OneHotEncoder,TreeEnsembleRegressor,ZipMap
 com.microsoft;1;BiasGelu,Fake_FunctionOp,FastGelu,FusedGemm,FusedMatMul,Gelu,Trilu