Concat: Fails in TILE Layout

[Sudharsan-V]

Describe the bug While performing the concat operation, it is observed that PCC drops significantly when the input is in the TILE layout, whereas the operation performs as expected in the ROW_MAJOR layout.

To Reproduce Steps to reproduce the behavior:

1. Save the below snippet to a file

   import torch
   import ttnn
   from tests.ttnn.utils_for_testing import assert_with_pcc
   def test_unit_permute(device, reset_seeds):
   tensor_a = torch.rand(1, 1, 768)
   tensor_b = torch.rand(1, 576, 768)
   torch_output = torch.cat((tensor_a, tensor_b), dim=1)
   
   tensor_a_rm = ttnn.from_torch(tensor_a, device=device, dtype=ttnn.bfloat16, layout=ttnn.ROW_MAJOR_LAYOUT)
   tensor_b_rm = ttnn.from_torch(tensor_b, device=device, dtype=ttnn.bfloat16, layout=ttnn.ROW_MAJOR_LAYOUT)
   ttnn_output = ttnn.concat((tensor_a_rm, tensor_b_rm), dim=1)
   ttnn_output = ttnn.to_torch(ttnn_output)
   assert_with_pcc(torch_output, ttnn_output, 0.99)  # PCC = 0.9999956481098589
   
   tensor_a_tile = ttnn.from_torch(tensor_a, device=device, dtype=ttnn.bfloat16, layout=ttnn.TILE_LAYOUT)
   tensor_b_tile = ttnn.from_torch(tensor_b, device=device, dtype=ttnn.bfloat16, layout=ttnn.TILE_LAYOUT)
   ttnn_output = ttnn.concat((tensor_a_tile, tensor_b_tile), dim=1)
   ttnn_output = ttnn.to_torch(ttnn_output)
   assert_with_pcc(torch_output, ttnn_output, 1)  # PCC = 0.00072778873165853

2. Run the script using pytest <path/to/file>
3. PCC drops when the input is in TILE layout.

[vigneshkeerthivasanx]

cc @saichandax @boris-drazic