#437 Uplift mlir to d542487 and update toHost to expect vector return type (#438)

### Ticket
#437 

### Problem description
`tt::runtime::toHost` was updated in mlir commit 10c88b1 to return
vector of tensors, which breaks torch build

### What's changed
Updated to use first tensor returned in list of tensors by toHost

### Checklist
- [x] New/Existing tests provide coverage for changes

Author: brataTT

third_party/CMakeLists.txt

@@ -3,7 +3,7 @@
 # SPDX-License-Identifier: Apache-2.0
 #
 
-set(TT_MLIR_VERSION "2d38c0af771c191ff2fe34469926e5c455016ffd")
+set(TT_MLIR_VERSION "d5424879e434711388f17cbc881e437b0b84ca1f")
 set(TORCH_MLIR_VERSION "a265d283357f572c5437f9217ea85fa9770d374c")
 
 if (BUILD_TOOLCHAIN)

tt_torch/csrc/bindings.cpp

@@ -109,7 +109,7 @@ std::vector<int64_t> as_vec_int64(std::vector<T> const &vec) {
 static torch::Tensor create_torch_tensor(const tt::runtime::Tensor &tensor,
                                          const tt::runtime::TensorDesc &desc) {
   tt::runtime::Tensor untilized_tensor =
-      tt::runtime::toHost(tensor, /*untilize=*/true);
+      tt::runtime::toHost(tensor, /*untilize=*/true)[0];
   const std::vector<std::int64_t> shape = as_vec_int64(desc.shape);
   const std::vector<std::int64_t> stride = as_vec_int64((desc.stride));