Remove obsolete workarounds.

[pp-mo]

Closes #239

I'm happy that none of these workarounds are now needed, as all fixed in eccodes. Long ago, I think.
This passes existing tests, but there is no effective testing of the relevant code.
What I have done is a roundtrip test : that at least does seem to work :

>>> ellipsoid = GeogCS(EARTH_RADIUS)
>>> data = np.arange(12).reshape(3, 4)
>>> cube = Cube(data, "air_pressure_anomaly")
>>> trans_merc = TransverseMercator(
...     49.0, -2.0, -400000.0, 100000.0, 0.9996012717, ellipsoid
... )
>>> coord = DimCoord(
...     np.arange(3),
...     "projection_y_coordinate",
...     units="m",
...     coord_system=trans_merc,
... )
>>> cube.add_dim_coord(coord, 0)
>>> coord = DimCoord(
...     np.arange(4),
...     "projection_x_coordinate",
...     units="m",
...     coord_system=trans_merc,
... )
>>> cube.add_dim_coord(coord, 1)
>>> coord = DimCoord(
...     np.arange(1),
...     "time",
...     units="days since 2000-01-01",
... )
>>> cube.add_aux_coord(coord, ())
>>> 
>>> print("\nInital cube to save:"); print(cube)

Inital cube to save:
air_pressure_anomaly / (unknown)    (projection_y_coordinate: 3; projection_x_coordinate: 4)
    Dimension coordinates:
        projection_y_coordinate                             x                           -
        projection_x_coordinate                             -                           x
    Scalar coordinates:
        time                        2000-01-01 00:00:00
>>> print("\nsaved + reloaded ...\n")

saved + reloaded ...

>>> result = iris.load_cube('temp.grib2')
>>> print(result); print("  ==> re-loaded coord system = ", result.coord_system())
unknown / (unknown)                 (projection_y_coordinate: 3; projection_x_coordinate: 4)
    Dimension coordinates:
        projection_y_coordinate                             x                           -
        projection_x_coordinate                             -                           x
    Scalar coordinates:
        forecast_period             0 hours
        forecast_reference_time     2000-01-01 00:00:00
        time                        2000-01-01 00:00:00
    Attributes:
        GRIB_PARAM                  GRIB2:d255c255n255
  ==> re-loaded coord system =  TransverseMercator(latitude_of_projection_origin=49.0, longitude_of_central_meridian=-2.0, false_easting=-400000.0, false_northing=100000.0, scale_factor_at_central_meridian=0.9996012449264526, ellipsoid=GeogCS(6371229.0))
>>> 

[stephenworsley]

_save_rules also defines fixup_int32_as_int32 and fixup_int32_as_uint32, should these also be removed likefixup_float32_from_int32? Their docstring also mentions how they are a workaround, is this for the same issue?

[pp-mo]

Thanks @stephenworsley
Now rebased, and removed the remaining "fixup" routines which you mentioned -- which were now completely unused, anyway

[stephenworsley]

Looks good, there's a couple comments which ought to be removed and this should be good to go.

[pp-mo]

Thanks @stephenworsley good spots !
I think I addressed those 3 now, please take another look.

[stephenworsley]

Looks good.

[pp-mo]

Thanks @stephenworsley !