general processor improvements

[johntruckenbrodt]

This improves on the general processing workflow stability:

• [dem.mosaic] the progress bar has been removed, which otherwise clutters logfiles in batch jobs
• a new configuration parameter snap_gpt_args enables parametrization of the underlying SNAP GPT call; this has been found particularly relevant in an HPC environment where SNAP may try to use more resources than ordered in the SLURM job; all functions of the S1_NRB.snap have been extended with an argument gpt_args accordingly
• snap.process: a new internal flag utm_multi enables processing to a single UTM zone if set to False; this is only intended for debugging and experimentation purposes
• preprocessed GRDs are no longer deleted so that they can be reused for buffering of neighboring scenes; a mechanism to delete these files once they are no longer needed is yet to be implemented (closes Keep pre-processed GRD files for reuse #75, new issue: delete preprocessed GRD products #86)
• GRD buffering: the method has been improved to no longer buffer a fixed number of source pixels but rather a multiple of the target pixel spacing in meters; the former might be insufficient if the target spacing is lower than the source spacing, in particular in combination with the edge cleaning performed during snap.postprocess

further bug fixes:

• metadata.extract.calculate_geolocation_accuracy: raise an additional error if the ellipsoidal incidence angle image only contains 0 (nodata)
• nrb.get_datasets: raise an error if no processing output could be found but just delete the scene from the list without raising an error if the files do not contain valid data
• metadata.extract.calc_performance_estimates: fixed bug in noise power file name parsing
• processor.main: if the scene footprints are used as geometry (aoi_tiles and aoi_geometry undefined), ensure full tile coverage at the ends of the data take groups

[johntruckenbrodt]

I have a hunch that with the latest changes (or even before that) the function nrb.format and the underlying nrb.get_datasets don't ensure that incomplete tiles are created. This needs to be checked before merging this PR. Will get to it next week. @maawoo, FYI.