tt_lib.tensor.bmm and matmul - inconsistent documentation

[nemanjagrujic]

For tt_lib.tensor.bmm and tt_lib.tensor.matmul ops:

The documentation does not state that operands need to be on device. But when operation is run with inputs in system memory we get error:

input_tensor_a.storage_type() == StorageType::DEVICE and input_tensor_b.storage_type() == StorageType::DEVICE
info:
Operands to matmul need to be on device!

[bbradelTT]

tt_lib bmm/matmul no longer exist and everything in python goes through ttnn matmul/linear. I updated the documentation for the latter to indicate that the tensors need to be on the device.