Support FullSumState in `VMC_SR` and `Infidelity_SR`

[PhilipVinc]

The 'hard' part of the implementation is inserting reweighing terms everywhere in srt.py and srt_ntk.py.
Luckily we already have a version of those with the reweighing terms as part of our IS paper

https://github.com/NeuralQXLab/importance_sampling_nqs/blob/main/packages/advanced_drivers/_src/driver/ngd/srt_onthefly.py#L40

https://github.com/NeuralQXLab/importance_sampling_nqs/blob/main/packages/advanced_drivers/_src/driver/ngd/sr_srt_common.py

where 'weights' are assumed to sum up to 1.

Implementing this should be 'easy' in the sense that the only thing that needs to be done, apart from pasting the code I linked ubove inside netket, is to pipe down the driver.state.probability_distribution to the SR logic.

If someone wants to take a stab at it, I can give more pointers.