Pass device to Hamiltonian, pass to pauli_string_to_matrix

[yannick-couzinie]

pauli_string_to_matrix takes a device argument, which defaults to torch.device('cpu'):

torchquantum/torchquantum/util/utils.py

Line 990 in 611cc2a

def pauli_string_to_matrix(pauli: str, device=torch.device('cpu')) -> torch.Tensor:

When trying to use Hamiltonian on GPU this invariably leads to device errors, since the result from pauli_string_to_matrix will be on CPU, so it would be nice to be able to pass device to Hamiltonian, which passes it on to pauli_string_to_matrix. Setting the default to None ensures that not previous behaviour is changed.