PowerLaw raises an error if sampling interval contains negative values

[j-fletcher]

Description

Obtaining a sample $x$ from a power law distribution of degree $n$, defined over the interval $[a, b]$, requires evaluating the following:

$x = (a' + \xi\times\Delta)^{1/(n+1)}$,

where the offset $a' = a^{n+1}$, the span $\Delta = b^{n+1} - a'$, and $\xi$ is a (pseudo)random number sampled from $U[0, 1]$. In OpenMC, this operation uses numpy.power() when performed in Python and std::pow() in C++, both of which raise an error if a negative number is raised to a non-integer power.

Currently, there are no warnings against specifying an interval wherein $a' + \xi\times\Delta < 0$ for some value of $\xi$, and when std::pow() is called in such cases, OpenMC raises the following RuntimeError:

RuntimeError: Too few source sites satisfied the constraints (minimum source rejection fraction = 0.05). Please check your source definition or set a lower value of Settings.source_rejection_fraction.

This PR checks the values of $a$ and $b$ when a PowerLaw is created to make sure that the distribution is defined over a nonnegative-valued interval. This makes clear the cause of the above error before it occurs.

Fixes #3541

Checklist

• I have performed a self-review of my own code
• I have run clang-format (version 15) on any C++ source files (not applicable)
• I have followed the style guidelines for Python source files (if applicable)
• I have made corresponding changes to the documentation (if applicable)
• I have added tests that prove my fix is effective or that my feature works (not applicable)

[eepeterson]

Thanks for the added checks, to be consistent with the documentation, it might be worth adding an explicit check that b > a in the constructor so that we're positive that a is the lower bound and b is the upper bound. It doesn't actually change the distribution of sampled values if they are flipped I don't think, but it could be confusing.

[j-fletcher]

Good idea--I've made that change in the PowerLaw constructor now.

There requested changes are already fixed and another reviewer accepted them.