[Triton-MLIR] Fix threadsPerWarp derivation in BlockedEncodingAttr (#722)

Example:

```
    auto encoding = triton::gpu::BlockedEncodingAttr::get(
        &getContext(), {8, 32}, {2, 2}, {1, 0}, 2);
     //shape = [32 x 8], order = [1, 0], sizePerThread=[2, 2], numWarps=2
```

Expected output:

```
      //#triton_gpu.blocked_layout<{
      //  sizePerThread = {2, 2}
      //  threadsPerWarp = {8, 4}
      //  warpsPerCTA = {2, 1}
      //}>
```

Incorrect output by the current branch

```
      //#triton_gpu.blocked_layout<{
      //  sizePerThread = {2, 2}
      //  threadsPerWarp = {16, 2}
      //  warpsPerCTA = {2, 1}
      //}>
```

include/triton/Dialect/TritonGPU/IR/TritonGPUAttrDefs.td

@@ -169,6 +169,7 @@ for
         int dim = order[_dim];
         int maxNumThreads = int(shape[dim]) / sizePerThread[dim];
         warpsPerCTA[dim] = std::clamp(remainingWarps, 1, maxNumThreads);
+        maxNumThreads = maxNumThreads / warpsPerCTA[dim];
         threadsPerWarp[dim] = std::clamp(remainingLanes, 1, maxNumThreads);
         remainingWarps /= warpsPerCTA[dim];
         remainingLanes /= threadsPerWarp[dim];