Erasing casting all floats on input and output to float32

[dgolubovicTT]

Erasing these lines since they cast all tensors from any float to float32. This bug was hidden in pybuda because we hadn't used third return value of the method
translated_modules, translated_device_types, _ = generate_pybuda_module(module, pt_inputs, compiler_cfg, module.name, verify_cfg)

This third parameter is actualy translated_inputs, which had its data format changed by tvm in these lines.

@nvukobratTT do you maybe know why we did this? It doesn't seem desirable.

[nvukobratTT]

@dgolubovicTT can't recall for sure. But I agree that this isn't a valid case as it'll tread all float types as float32 (e.g. float16).