some cleaning of screen5

[zingale]

this will change answers to roundoff because we do sqrt(sqrt(x)) instead of pow(x, 0.25)

[zingale]

here are the tests:
http://groot.astro.sunysb.edu/Microphysics/test-suite/gfortran/2025-12-16-001/index.html

if you look at the screening test, you can see that the sqrt(sqrt(x)) change introduces roundoff diffs.

http://groot.astro.sunysb.edu/Microphysics/test-suite/gfortran/2025-12-16-001/test_screening_C.html

For a few tests, the burning takes an additional few steps, giving seemingly large diffs in the output. But I think that the diffs are okay.

We might be able to write our own qdrt() function that does the automatic differentiation more efficiently.