Goodness-of-Fit Scoring, QOL, and Functional Updates

[xia-fr]

New Content:

• Ported goodness-of-fit (GoF) scoring system from Seismosoil (Matlab version) to PySeismoSoil.
• Batch calculation for GoF scoring.

Functional Changes:

• Added warning for unrealistic rho calculation due to unrealistic overburden pressure for surface layer if it is too thin.
• Rho calculation now uses the depth of the midpoint of layers.
• Filter kernel for signal processing now uses the recommended 'sos' output type.
• Added option in batch simulation class to capture ValueErrors, so a single erroneous simulation doesn't interrupt a batch run.
• Added option in batch simulation class to use the forkserver context of spawning subprocesses (for Unix systems). This is recommended to prevent slow code in use cases where additional modules and variables other than the ones associated with PySeismoSoil are present in a code.

Addressing Errors:

• For nonlinear simulations, the dummy curves for ggmax and xi are now created separately from their corresponding calibrated parameters. The values contained in these dummy curves are not consequential for the simulations, but previously both the dummy ggmax and dummy xi curves were generated off of only the ggmax parameters. This would occasionally trigger a ValueError related to unrealistic damping values, since the ggmax parameters are only calibrated to guarantee the shape of the ggmax curve.
• For hh calibration, added handling for edge cases because the algorithm searches extremely large or small values of 'a'. This prevents RuntimeWarning due to extremely large/small exponentials.

QOL:

• Multiple curve plotting (in class_curves) now plots all curves on one plot, instead of creating a new plot for every curve.
• If verbose=False, the outputs from the nonlinear NLHH executables and the other simulation classes are fully suppressed. This prevents flooding of the console when running 1000+ simulations in batch.