Analytical gamma second derivative

[bhourahine]

Replace finite difference for global hybrids.

[vanderhe]

@bhourahine Did you already check that the new routines reproduce the old ones? This is not necessarily caught by the regression tests.

[vanderhe]

Style guide violations in most lines ;)

Suggested change

	& (2.0_dp*tauA**3/48.0_dp + 2.0_dp/dist**3 - tauA*(2.0_dp*dist*tauA**3/48.0_dp&
	& (2.0_dp * tauA**3 / 48.0_dp + 2.0_dp / dist**3 - tauA * (2.0_dp * dist * tauA**3 / 48.0_dp&

+induced line breaks, if applicable

[bhourahine]

@bhourahine Did you already check that the new routines reproduce the old ones? This is not necessarily caught by the regression tests.

getddYGammaSubPart was explicity tested, but not the full term. If there is no coverage, how was the original expression tested? And any reason why a test can't be added?

[vanderhe]

@bhourahine Did you already check that the new routines reproduce the old ones? This is not necessarily caught by the regression tests.

getddYGammaSubPart was explicity tested, but not the full term. If there is no coverage, how was the original expression tested? And any reason why a test can't be added?

Back then I verified that the 5-th order polynomial smoothly connects and plotted the whole thing. Although there is at least one test with the damped TCI, I'm not sure if it really hits the damped region.