reduce_window fusions

[avik-pal]

      %7 = stablehlo.slice %6 [0:511, 1:511] : (tensor<512x512xf32>) -> tensor<511x510xf32>
      %8 = "stablehlo.reduce_window"(%7, %cst) <{window_dimensions = array<i64: 2, 1>}> ({
      ^bb0(%arg2: tensor<f32>, %arg3: tensor<f32>):
        %26 = stablehlo.add %arg2, %arg3 : tensor<f32>
        stablehlo.return %26 : tensor<f32>
      }) : (tensor<511x510xf32>, tensor<f32>) -> tensor<510x510xf32>
      %9 = stablehlo.slice %6 [2:512, 1:511] : (tensor<512x512xf32>) -> tensor<510x510xf32>
      %10 = stablehlo.add %8, %9 {enzymexla.symmetric_matrix = [#enzymexla<guaranteed NOTGUARANTEED>]} : tensor<510x510xf32>

reduce and elementwise fusions already handle chains nicely. we should be fusing these ops into a larger reduce window