Update to latest SUNDIALS v7.6.0

[rpep]

Some changes to underlying library but nothing major, tests pass locally (fingers crossed for CI!)

[davidcortesortuno]

Is there any improvement in updating to Sundials 7.x ? From your commits the data types in the cvode pyx file look much cleaner :O

[rpep]

@davidcortesortuno not massively from what I can tell, just trying to stay relatively modern!

I've bumped fftw too, I had to remove the finalize though as it crashes with it when you reinit fftw. There's not really a good way to fix this properly, other than have a cleanup that only ones once in a singleton or something :/ The only real benefit to running finalize is that if you were for e.g. running a simulation that had Demag and then removed it, you'd have more memory available for the second stage, but I think it's probably unnecessary for this.

[davidcortesortuno]

Awesome. Regarding the fftw cleanup function, I guess we could wrap it and call it only if the user wanted to get rid of all created demag memory. Because if you create a new demag, my guess is that all fftw plans are created again?

[rpep]

@davidcortesortuno we already call fftw_destroy_plan, this is more like shared allocated global stuff across plans that gets cleaned up. It's absiclay only useful if you need full memory back after finishing all FFTs but if you are just ending the program obviously everything will be deallocated anyway.

[davidcortesortuno]

Indeed, yes. Also we never use more than one demag most of the time, so probably full deallocation is not that useful. I'll merge this PR if everything is passing.