[mlir][linalg] Decompose winograd operators (#96183)

Convert Linalg winograd_filter_transform, winograd_input_transform, and
winograd_output_transform into nested loops with matrix multiplication
with constant transform matrices.

Support several configurations of Winograd Conv2D, including F(2, 3),
F(4, 3) and F(2, 5). These configurations show that the implementation
can support different kernel size (3 and 5) and different output size
(2 and 4). Besides symetric kernel size 3x3 and 5x5, this patch also
supports 1x3, 3x1, 1x5, and 5x1 kernels.

The implementation is based on the paper, Fast Algorithm for
Convolutional Neural Networks. (https://arxiv.org/abs/1509.09308)

Reviewers: ftynse, Max191, GeorgeARM, nicolasvasilache, MaheshRavishankar, dcaballe, rengolin

Reviewed By: ftynse, Max191

Pull Request: #96183

Author: Hsiangkai

mlir/include/mlir/Dialect/Linalg/Transforms/Transforms.h

@@ -1746,6 +1746,9 @@ void populateBlockPackMatmulPatterns(RewritePatternSet &patterns,
 void populateWinogradConv2DPatterns(RewritePatternSet &patterns, int64_t m,
                                     int64_t r);
 
+/// Patterns to decompose Winograd operators.
+void populateDecomposeWinogradOpsPatterns(RewritePatternSet &patterns);
+
 /// Adds patterns that reduce the rank of named contraction ops that have
 /// unit dimensions in the operand(s) by converting to a sequence of `collapse_shape`,
 /// `<corresponding linalg named op>`, `expand_shape` (if on tensors).  For example a