Fixes #445: GRIB1 data length overflow

[cofinoa]

Due to a trick done by ECMWF's GRIBEX to support large GRIBs, we need a special treatment to fix the length of the GRIB message. See:
https://software.ecmwf.int/wiki/display/EMOS/Changes+in+cycle+000281
#445

[cofinoa]

I'm not sure if I have made the PR on the correct branch.
The commit has been made for the the tag 4.6.11.

[lesserwhirls]

Nice - thank you for the contribution, @cofinoa. Would you happen to have a record or two that we could use for tests? This is something that we'd want a test, for sure.

edit: I see @sefadolunay provided a few sample grids on that issue, so we're good there.

[lesserwhirls]

I'm currently running this PR on our Jenkins instance to see how our full tests turn out.

[cofinoa]

I have extracted one GRIB message from @sefadolunay sample file:
https://meteo.unican.es/droppings/antonio/ecmwfGRIB1large.grib

[lesserwhirls]

Our test suite on Jenkins passed without any impact on our GRIB tests, so that's good. It would be good to get this on 5.0 as well, and I think the changes should go on cleanly. I'll add some tests for this on 5.0 once we get it applied there. @cofinoa - would you mind rebasing your PR to pick up the latests changes in master?

[cofinoa]

@lesserwhirls

I have rebased my PR to the latest changes at master ... or at least I think I have made ...

[sefadolunay]

Here I provide more test files: test1 test2