Handle case when granules from the same collection have different variables

[wveit]

As an example, the MODIS_T-JPL-L2P-v2019.0 collection has some granules that contain the variables

• quality_level_4um
• sea_surface_temperature_4um
• sses_bias_4um
• sses_standard_deviation_4um
  while other granules do not.

Currently, concise is giving the error message "Variable schemas are inconsistent between granules" when attempting to concatenate such granules.

One solution might be to create the missing variables for granules that don't have them, and fill it with the empty fill value, before merging the granules.

Related to #12