iMS100 GDP: NaN mismatch for 'val' and 'uc_tot'

**Describe the bug**
The error message `NaN mismatch for 'val' and 'uc_tot'` is triggered when GDP measurements have a NaN uncertainty.

It occurred for #205 already, but now it appears that that the iMS100 GDP also regularly has a handful of measurement points with no uncertainty for the wind direction and wind speed.

For now, dvas will identify them, flag them as "invalid", and mask them when assembling the CWS. But this really should be fixed/handled at the GDP creation level.

**Edit:** following #260, it appears that some points can also have an uncertainty of 0 for the `wdir` and `wspeed` of the iMS100. These are now also flagged as invalid by dvas, until they can be fixed at the GDP level. Having 0 total uncertainty wreaks havoc when normalizing things.