[Bug Report] Add op does not respect provided output memory config

[odjuricicTT]

Describe the bug When passing a sharded memory config into add op, core_gird attribute is overridden on op execution.

To Reproduce

import ttnn

device_id = 0
device = ttnn.open_device(device_id=device_id)

ttnn.SetDefaultDevice(device)

try:
    tensor = ttnn.zeros((128, 32))
    tensor = ttnn.to_layout(tensor, ttnn.TILE_LAYOUT)
    tensor = ttnn.to_device(tensor, device, memory_config=ttnn.MemoryConfig(ttnn.TensorMemoryLayout.INTERLEAVED, ttnn.BufferType.DRAM))

    out_mem_config = ttnn.create_sharded_memory_config(
        shape=(128, 32),
        core_grid=ttnn.CoreGrid(x=2, y=1),
        strategy=ttnn.ShardStrategy.HEIGHT,
        orientation=ttnn.ShardOrientation.ROW_MAJOR,
    )

    print("Desired grid:", out_mem_config.shard_spec.grid)
    # Desired grid: {[(x=0,y=0) - (x=1,y=0)]}
    out = ttnn.add(tensor, tensor, memory_config=out_mem_config)

    print("Output grid:", out.memory_config().shard_spec.grid)
    # Output grid: {[(x=0,y=0) - (x=3,y=0)]}
finally:
    ttnn.close_device(device)

Expected behavior The output is sharded on the core_grid that was provided or an error is thrown if that is not possible.

Please complete the following environment information:

• OS: Ubuntu 22.04.5
• Commit: ca03e4d9ac0d44803e3ed761c427fb762da4794c

[odjuricicTT]

Hey @eyonland, did you have time to take a look at this?