Handling of Fx variables as regular variables for CMIP5 and CMIP6 (simple, no CMOR checks)

[valeriupredoi]

The simplified version of #1037 (there are no CMOR checks and fixes for fx vars in this PR) - basically uses the standard input and output data structures for FX variables as well, removing the need for special fx I/O; tested with CMIP5 and CMIP6 data; note that now if a variable needs fx files for diagnostic they are specified slightly more complex:

    variables:
      tos:
        preprocessor: pp_rad
        mip: Omon
        grid: gn
        fx_files: [{short_name: areacello, mip: Ofx}, ]

since for CMIP5 this was not needed because mip=fx always for fx variables in CMIP5; a bit different for CMIP6 though.

@bouweandela can u pls have a look at the recipe tests that are systematically failing? I fail to see why this is happening. Cheers, dude 🍺

[valeriupredoi]

OK guys, compacted code, fixed all tests, tested extensively with CMIP5 and 6 -> this is ready for your consideration 🍺 @bouweandela @zklaus @mattiarighi

[valeriupredoi]

whereas #1037 can wait (it's big and ugly 😁 ) I reckon this one should have some priority - people are already running CMIP6 data that needs fx files (like @tillku ) - when you have time @bouweandela and @mattiarighi 🍺

[bouweandela]

Please re-open this pull request in the ESMValTool Core repository. You can do so by pushing this branch to that repository, using the following instructions.

First add the repository with

git remote add esmvalcore git@github.com:esmvalgroup/esmvalcore

or

git remote add esmvalcore https://github.com/esmvalgroup/esmvalcore

if you're not using ssh to connect to GitHub.

Next upload the branch with

git push esmvalcore your_branch

and open a pull request here

[valeriupredoi]

ported to ESMValGroup/ESMValCore#18