SEVIRI L1b netCDF reader not dask-compliant

[sjoro]

seviri_l1b_nc-reader is not using dask, i.e. get_dataset-method returns a plain xarray.Dataset-object.

The reader should be "daskified", which may have an impact on SEVIRICalibrationHandler in seviri_base.py and other readers that use this class.

[BENR0]

Not directly related to dask but when working on the reader the following should probably be changed too. I noticed that the nc reader uses mask_and_scale=False while opening the dataset. These attribute (among others) are not removed from the dataset and therefore lead to what I think are unwanted results when writing out data with the cf writer. I think some of the attributes of the original file should be stripped (or updated) because they don't make sense later on, or are confusing (e.g. comment, long_name, nc_key, valid_min, valid_max, etc.).