[emu-sv] sampling more than 25 atoms - multinomial restricts this

[murogrande]

running the following code:

from emu_sv import StateVector

qubit_count = 26
eigenstate = ('r','g')
amplitudes = {"r"*qubit_count:1.0}

state = StateVector.from_state_amplitudes(eigenstates=eigenstate,amplitudes=amplitudes)
print("sv  sampling",state.sample())

I got the following error:

 outcomes = torch.multinomial(probabilities, num_shots, replacement=True)
RuntimeError: number of categories cannot exceed 2^24

• multinomial has an internal data type that restricts the number of categories to $2^{24}$.

We should have our own implementation since emu-sv simulate can more than 24 atoms