Fix several PIO warnings: guard empty-string attributes and match start/count dims to variable rank

[ndkeen]

Fix a collection of PIO warnings emitted when writing NetCDF attributes with empty
strings and when start/count arrays do not match a variable's actual rank. Changes
span EAM, ELM, MPAS-framework, and the MCT/MOAB drivers.

• Guard empty-string attribute writes (units, standard_name): wrap pio_put_att
  calls in if (len_trim(...) > 0) checks in driver-mct/seq_io_mod.F90,
  driver-moab/seq_io_mod.F90, cam_grid_support.F90, ncdio_pio.F90.in,
  restFileMod.F90, and mpas_stream_manager.F to prevent PIO from writing
  zero-length string attributes.

• Match start/count rank to variable rank (tracer_data.F90, phys_prop.F90):
  query pio_inq_varndims before pio_get_var; pass strt(1:1)/cnt(1:1) for
  1-D variables to eliminate "count array size does not match variable rank" warnings.

• Fix avgflag restart layout (cam_history.F90): add a dedicated avgflag_len
  string-length dimension and store avgflag as a 3-D variable to satisfy PIO's
  leading-dimension convention. Read path detects old 2-D layout for backward
  compatibility.

• Fix date/time string dimension in ELM history (histFileMod.F90): introduce
  hist_date_str_len = 8 for date_written/time_written dimensions instead of
  reusing the 16-character field-name dimension; widen scale-type variables from
  character(len=8) to character(len=hist_dim_name_length).

[BFB]

[ndkeen]

I started this effort with just a handful of cases, but then went through all of the tests in e3sm_integration to tackle the bulk of PIO warnings as it was clear some changes were impacting many cases.. I can break this up into components if it helps.

[jayeshkrishna]

@ndkeen : Regarding the last change (autodetect ELM input file type), I think instead of reading a file to determine if its NetCDF4, a better approach is to convert NetCDF4 input files to ones that PnetCDF can work with ("64-bit data"). The choice of I/O type to read the file needs to be made in the config/xml stage.
(I can work on the conversion if its too much work for you, but I would recommend dropping that commit)

[ndkeen]

Yea I was also suspicious of that change. I reverted.

[rljacob]

Maybe the model should error if it finds a NetCDF4 input file? That would encourage people to do the conversion!

[ndkeen]

OK to merge or do we want changes?

[jayeshkrishna]

@jonbob / @peterdschwartz : Please review the PR when you get a chance (please make sure that you are not losing any attributes. IMO, instead of removing the attributes a better approach would be to assign the correct value and write them out)

[peterdschwartz]

Looks ok for the elm files

[rljacob]

@jonbob please review.

[jonbob]

approved based on visual inspection and developer testing