add finalize subroutine for stochastic physics

[pjpegion]

This a new feature to add a finalize subroutine to deallocate array at the end of a run. Associated with PRs
NOAA-EMC/ufsatm#199 and
ufs-community/ufs-weather-model#278

[lisa-bengtsson]

Looks good to me.

[HenryRWinterbottom]

I would strongly suggest adding if-def type statements to this PR. For example, instead of:

deallocate(variable)

do:

if(allocated(variable)) deallocate(variable)

This better protects the code from possible and difficult to chase segmentation faults in the future -- at least based on my experience.

[pjpegion]

@HenryWinterbottom-NOAA good point, I thought about doing it, but was just lazy.

[HenryRWinterbottom]

@pjpegion If you are planning to do it, just send me a note (or submit a new PR) and I am happy to approve.

[pjpegion]

@HenryWinterbottom-NOAA I added the if (allocated()) conditionals

[HenryRWinterbottom]

Thank you @pjpegion.

Approved.

[climbfuji]

The PR looks okay; there are several places where a comment line

! Copy blocked data into contiguous arrays; no need to copy weights in (intent(out))

was moved around to places where it doesn't belong. I recommend removing or changing them, but certainly not a show stopper.

I am currently creating new baselines on cheyenne.intel, cheyenne.gnu and gaea.intel. The new baseline date tag will be 20201127, please update RTPWD in rt.sh accordingly.

[pjpegion]

Looks like the changes you are suggesting are for the fv3atm PR.

[climbfuji]

Looks like the changes you are suggesting are for the fv3atm PR.

Yes, sorry.