Efficient calculation of probabilities

[JacobHast]

Before posting a feature request

• I have searched exisisting GitHub issues to make sure the feature request does not already exist.

Feature details

The calculation of probabilities in lab_dev for states in the Fock representation appears relatively slow - at least it can be sped up by a factor 10 using a simple numpy code as shown in this example.

Implementation

No response

How important would you say this feature is?

2: Somewhat important. Needed this quarter.

Additional information

While the calculation of the probability of a single state is quite fast, I sometimes need to do it for >100,000 states, at which point this step becomes the numerical bottleneck, so it would be very useful to get this improved.
I have not done the comparison for DM states, only Ket

[JacobHast]

This also directly affects the speed of .normalize

[JacobHast]

Code also testing .to_fock() before .probability - doesn't improve things

import mrmustard.lab_dev as mm
import numpy as np
from mrmustard import settings
import timeit

settings.AUTOSHAPE_MAX = 1000

def make_state() -> mm.Ket:
    state = mm.Ket.from_fock([0], np.random.random(settings.AUTOSHAPE_MAX))
    return state

STATE = make_state()

def probability_np() -> float:
    state_fock = STATE.fock(settings.AUTOSHAPE_MAX)
    return np.sum(np.abs(state_fock) ** 2)

def probability_buildin() -> float:
    return STATE.probability

def probability_buildin_to_fock() -> float:
    return STATE.to_fock().probability

%timeit probability_np() # 22.4 µs
%timeit probability_buildin() # 218 µs
%timeit probability_buildin_to_fock() # 371 µs


assert np.isclose(probability_np(), probability_buildin())
assert np.isclose(probability_np(), probability_buildin())