Plev compile fix

[sit23]

As per issue #105, it was noted by users in Bristol that the compile_plev_interpolation.sh script was giving errors. I found that the reason was that it was trying to use the mkmf template from the postprocessing/bin directory, rather than the mkmf template used to compile the model. I had fixed this problem on the pre_ictp_mods branch back in August 2018, but hadn't merged it in. So apologies for that.

I have now updated the compile script so that it picks the same mkmf template as the model uses, and I've also added a simplified version of my run_plevel.py script, which was also taken from the pre_ictp_mods branch. There is no need to run the trip test in this case, as no fortran modifications have been made.

[sit23]

As we discussed this yesterday, maybe @wseviour could do the code review for me?

[sit23]

I've also added a change to the script created for removing restart files for data clearup.

[sit23]

Would one of @RuthG, @penmaher or @wseviour mind reviewing this for me so that I can merge it in? Nothing very substantial.

[ntlewis]

Just a quick note on this. In Oxford I'm currently using the -Duse_netCDF3 option when compiling Isca with Intel compilers. This is a temporary snag deriving from the fact that here the netcdf library compatible with Intel 2018 compilers was compiled without netCDF4 enabled.

I just thought I'd mention that the above won't work with a mkmf template that uses-Duse_netCDF3, as pinterp_utilities.F90 and run_pressure_interp.F90 assume netCDF4 and don't have any #ifdef statements to switch to a 'netCDF3' mode.

I can get around this by compiling the plevel interpolation stuff with gfortran (for which I can access a compatible netCDF4 library).