Request error handling improvements in ecCodes API from ECMWF

[Ian Culverwell]

ROPP-10.0 beta reviewers say

Using a plain fprintf() statement rather than a call to logging facility of the
library is, in our view, a bug in the upstream eccodes library. Have you
discussed the problem with user support at ECMWF? We also believe that
controlling the behaviour through an environment variable is a bad design choice
because different applications linked to the eccodes library might have
different needs; requiring to set environment variables depending on which
application is run isn't a workable option for ordinary users. Instead, each
application should be able to set the handling of missing values (and the
consistency with bufrdc in general) through the API. We haven't found an API
function in the eccodes library offering such a function; we thus suggest a
feature request to ECMWF to add such an API to eccodes.

We should consider this - at a low priority.

[Ian Culverwell]

-> ROPP12.0 at low priority.