Accuracy of interpolation for Witt and Mao 1994 (i.e. FSPL) magnification

[rpoleski]

We have interpolation for elliptical integrals in pointlens.py for WM94. These grid of elliptical integral of the 3rd kind was set to get relative accuracy of 10^-4. Based on a poster by @josepsn at the 2024 Microlensing Conference, I've found that for u_0 = 1.736 and rho = 0.23563 the relative error is 7.7*10^-4. I think we should improve the accuracy of the interpolation.