Split tclean PSF generation into separate function

[ddobie]

In #2 we've found there's a high memory requirement in a single step - generating the PSF.

One possible workaround for this is to run this step separately in a short job on a high memory node. This should be easily achievable by running tclean twice:

1. tclean(..., niter=0, calcpsf=True) on a high memory node to generate the PSF and primary beam
2. tclean(...,niter=X, calcpsf=False) for imaging the rest of the data.

See tclean docs: https://casadocs.readthedocs.io/en/v6.5.6/api/tt/casatasks.imaging.tclean.html.

The full fix would also require tweaking slurm scripts in order to introduce dependencies etc.

[Elle-WANG]

This is not an issue anymore as we don't need huge memory in ozstar