Use correct input file for general aerosol optics in IFS-style drivers

[reuterbal]

This is a patch to the 1.6.1-beta branch. I will file a similar one against master in the future.

I noticed that results differed with the 49r1 namelist file between the IFS-style drivers and the "default" ecrad driver, while being the same with the 47r3 namelist file.
I managed to trace this to a different aerosol data file being used, when general_aerosol_optics is active.

This PR contributes:

• a fix to radiation_setup that chooses a different aerosol data file when general_aerosol_optics is active
• a fix to radiation_setup that disables the call to DRY_AEROSOL_MASS_EXTINCTION, which fails with a segfault. Since the aerosol setup is already performed during the input reading, it is redundant in the standalone context anyway.
• a build option that optionally allows to supply effective radii, seed and cloud fraction to RADIATION_SCHEME. This way we can ensure the same values are used per column, independent of the chosen block size, and thus obtain bit-identical results between blocked and non-blocked drivers.
• an addition to the test specification in CMake that uses cdo (if available) to compare outputs between ifs-drivers and the non-ifs driver (with net flux output)