Added gradient spectrum calculations

[FrankZijlstra]

This PR adds the gradient spectrum calculation as available in MATLAB Pulseq: https://github.com/pulseq/pulseq/blob/master/matlab/demoUnsorted/gradSpectrum.m

The implementation provides a fair number of parameters to control how the calculation is actually done (as we're not entirely sure what is the best way to go about this). The defaults should be sufficient to avoid peaks in the spectrum at the acoustic resonances of your scanner (and avoid damaging the gradient coils).

Basic usage (will plot the spectrums and also return them):

seq.calculate_gradient_spectrum()

Or including plotting the acoustic resonance ranges from a Siemens .asc file:

from pypulseq.utils.siemens.readasc import readasc
from pypulseq.utils.siemens.asc_to_hw import asc_to_acoustic_resonances

asc = readasc(asc_name)
seq.calculate_gradient_spectrum(acoustic_resonances=asc_to_acoustic_resonances(asc[0]))

[schuenke]

As I am missing appropriate example data, I was not able to check if the code works as expected and especially if the results fit to the Matlab results. Would be nice if you could check this on your own and confirm that it's true.

Maybe you can even add a figure of the comparison to this PR.

Having (some) tests would ofc also be nice, but due to the fact that most of the code is currently untested, this is probably nothing I can ask for 😀

Still, I added some comments that might help to improve the code style and the compatibility with (future) type checkers.

[FrankZijlstra]

Python output for seq.calculate_gradient_spectrum(max_frequency=10000, combine_mode='rss') (parameters to be compatible with MATLAB defaults):

MATLAB output:

The different Y-scale is due to scipy's spectrogram function using different units than the computations in the MATLAB script. For now the Y-scale is arbitrary anyway, until we can figure out reasonable limits for acoustic resonances.