Expand analysis of satellite-based inversions

[melodb]

Christoph Riess (Empa) tried fluxie to analyse his satellite-based inversion outputs. Here is some feedback of what could be improved:

Spatial plots:

• add map of observation density (number of observations per lat/lon grid cell)
• add map of observed/prior/posterior concentrations (similar to the map for prior/posterior fluxes)
• add option to set color scale to logarithmic

Concentration timeseries:

• allow slicing of the timeseries over a given region (by providing lat/lon limits or region name)
• compute mean/std of observed/prior/posterior concentration over that region
• plot this timeseries
• add timeseries of observation density (per region)

Add flexibility to fluxie:

• allow fluxes to be in any time frequency (instead of yearly/monthly/...)
• allow region definition directly on fluxie instead of relying on regions pre-defined in the flux output file (using user defined lat/lon limits or python country masks)