[standalone MPAS] Finalize PIO when finalizing MPAS i/o

[xylar]

This allows SCORPIO to print out timing information needed for some types of debugging.

pio_finalize() is only called in standalone. If MPAS components are run as part of an E3SM run, the PIO context is passed to MPAS and this signals that the MPAS framework should not finalize PIO.

[xylar]

See discussion PR E3SM-Ocean-Discussion#30 for additional context.

[xylar]

Testing

I ran the compass ocean pr test suite and verified that the expected io_perf_summary_*.txt and io_perf_summary_*.json files are produced for each forward run of MPAS-Ocean. These files were previously not available.

I ran the ERS.ne11_oQU240.WCYCL1850NS.anvil_intel E3SM test and verified that timing appears as usual in the spio_stats/ subdirectory (presumably a good indication that the MPAS framework didn't erroneously finalize PIO too soon).

[xylar]

@jonbob, would you like me to do further E3SM testing on this to make sure I'm doing no harm? Happy to run whatever tests you suggest.

[jonbob]

@xylar - I'll do more testing before I merge, so I think this looks fine. Thanks

[mark-petersen]

This functions for the ocean core. Tested with gnu and intel, both debug and optimized, nightly suite on badger. Also compares bfb against branch point with gnu optimized.

[matthewhoffman]

Approving based on discussion yesterday, inspection, and Mark's testing. This appears to be the proper implementation that mirrors the init operations.

[xylar]

Thanks for these reviews as well, @matthewhoffman and @mark-petersen! Much appreciated.

@akturner, this is lower priority than #5162 for me but would appreciate you having a look at this soon, too, if you can.

[xylar]

Thanks @akturner!

[jonbob]

Test merge passes sanity testing on chrysalis:

• SMS_D_Ld1.ne30pg2_EC30to60E2r2.WCYCL1850.chrysalis_intel.allactive-wcprod
• ERS.ne11_oQU240.WCYCL1850NS.chrysalis_intel

will merge to next as soon as the repo is open

[jonbob]

merged to next

[jonbob]

merged to master