Fix Papaevangelo_2010 friction factor: log10 not natural log in numerator

[gaoflow]

Papaevangelo_2010 computes the numerator term (7 - log Re)^4 with math.log (natural log), but the denominator and the original correlation use log10. The result is wrong across the method's registered range (Re 1e4-1e7, eD 1e-5-1e-3): up to ~260% error versus Colebrook, and negative friction factors above Re ~ 3e6.

>>> from fluids.friction import Papaevangelo_2010
>>> Papaevangelo_2010(1e7, 1e-4)
-0.019956280684600055   # a negative friction factor, inside the valid range

Switching the numerator to log10 drops the max error over the range to ~0.6%, matching the correlation's stated accuracy. I also corrected the docstring formula (it showed \ln) and the doctest / test_friction reference values, which were both baked from the buggy output. Added a range sweep asserting the correlation stays positive and within 1% of Colebrook across its domain.

[CalebBell]

Thank you! Nice catch. The CI failure show that the Colebrook(Re, eD, tol=0) call uses mpmath, but this test isn't masked as requiring mpmath so it tries to run on platforms without mpmath and fails. Colebrook is very high precision by itself. Can you remove the tol=0 argument? Otherwise this looks good to merge.