#2133: Modified existing golden functionality code to support multi-device tensors for CCL operations

[tapspatel]

Golden functionality currently only supports matching single device tensors. This change is step 1 to get the golden infrastructure ready to store per device tensors in the golden map as well as get the device tensors per device in runtime.

The following APIs are changed

op_golden_tensor_map = binary.get_debug_info_golden(loc)

{
device : golden_tensor
}

op_output_tensor_map = ttrt.runtime.get_op_output_tensor(op_context, program_context)

{
device: device_tensor
}

[vprajapati-tt]

Yo @tapspatel I think I have some conflicting changes in #2136, we should try and coordinate a strategy for this.

[tapspatel]

@vprajapati-tt no worries, feel free to merge yours and I'll rebase on-top of it. I approved your PR

[tapspatel]

keeping this PR open but in draft and this is not the approach we want to take right now