Avoid overflow errors with _get_radial_sum

[cmaloney111]

The _get_radial_sum function currently overflows for large autocorrelation values. This can lead to negative values for the probability in the two_point_correlation function.

Problem
When using the two_point_correlation function with large autocorrelation values, I encountered overflow errors:

• When using porespy through Anaconda, the overflow results in negative probability values (see attached screenshot).
• On my local machine (where I copied the two_point_correlation code), it causes an OverflowError with the message "Python int too large to convert to C long."

Solution
To fix this issue, I:

• Specified the dtype of radial_sum as float64 to handle large values without overflow.
• Added a check to avoid division by zero using if np.any(mask).

[jgostick]

Thanks for this. This reminds me of another bug I fixed recently, where something np.sum(im) would use the dtype of im when doing the sum. I fixed it by specifying the dtype in the sum function, similar to what you did here. Thanks for catching this.

[cmaloney111]

No problem!