Test case for GMI.

test data (5 minute granule):
  1C.GPM.GMI.XCAL2016-C.20180418-S212640-E213138.V05A.RT-H5

Shell script: run_gmi.sh
========================

This has 3 parts:

1) Convert hdf to BUFR using the standalone gmi_hdf2bufr.exe. Note that 2 BUFR
files are generated, one for each set of channels.

2) Run gmi_main.exe, with the namelist (gmi_main.nl) configured to perform
averaging and map to a user-defined grid.

3) Create quick-look images of the original and averaged data, using
mwipp_quicklook.py.

The python plotting script uses Cartopy: see https://scitools.org.uk/cartopy/docs/latest/

Input data file
1705715  1C.GPM.GMI.XCAL2016-C.20180418-S212640-E213138.V05A.RT-H5

Top-level namelists:
gmi_main.nl

Lower-level namelists:
gmi_averaging.nl, mwipp_bufr.nl, user_grid.nl

User grid file:
t511_n256_lsm.grib

Output data files:
1688509 1C.GPM.GMI.XCAL2016-C.20180418-S212640-E213138.V05A.RT-H5_S1.bufr
1009818 1C.GPM.GMI.XCAL2016-C.20180418-S212640-E213138.V05A.RT-H5_S2.bufr
1339868 1C.GPM.GMI.XCAL2016-C.20180418-S212640-E213138.V05A.RT-H5_generic_S1.h5
 736022 1C.GPM.GMI.XCAL2016-C.20180418-S212640-E213138.V05A.RT-H5_generic_S2.h5
  45436 1C.GPM.GMI.XCAL2016-C.20180418-S212640-E213138.V05A.RT-H5_grid.bufr
 117888 1C.GPM.GMI.XCAL2016-C.20180418-S212640-E213138.V05A.RT-H5_grid.h5

Output image files:
 286512 1C.GPM.GMI.XCAL2016-C.20180418-S212640-E213138.V05A.RT-H5_generic_S1.h5_1.png
 248666 1C.GPM.GMI.XCAL2016-C.20180418-S212640-E213138.V05A.RT-H5_grid.h5_1.png

Approximate run times:
Part 1: 0.6s (hdf to BUFR)
Part 2:  14s (hdf read; average; re-map; BUFR write; HDF5 write)

You can inspect your output data files using tools such as h5dump (part of hdf5)
and bufr_dump (part of ecCodes), and compare with equivalent outputs in the
"reference_products" directory.

You should expect the output brightness temperatures to be consistent with the
reference outputs, to approximately 0.01K.
