Bounds as inputs and chips outputs

[crangelsmith]

Fixes #8 and #25

• [ ] Move from 'coords' to 'bounds' in the pipeline input and main attribute of the Downloader and Procesor classes (bounds are list with 4 entries as left, bottom, right, top).
• [ ] Remove 'region_size' option as not used anymore
• [ ] After the tiff files are downloaded the bounds are read from the tiff files and used downstream (given a projection, this is now in eastings, northings instead of lat long).
• [ ] Processed images (chips and RGB images use the updated projected bounds in their naming convention).
• [ ] Added the date range in the output names, in order to provide all relevant info in the name.
• [ ] Chips are now identified with the left_bottom corner + chip dimension + date range used in the download.
• [ ] Count data is now stored at the chip level as a json file with summary statistics (mean, std, min,max,etc), it has same name as the chip array.
• [ ] Removed config files that do not follow the new nomenclature of bounds instead of coords.

[crangelsmith]

This is how the new output of the pipeline looks like:

Notice that the directory with the highest level still has the lat,long bounds, but all processed images are now in eastings and northings.

[crangelsmith]

As discussed:

• The bits I've added need tidying up 😱
• Before splitting into chips, confirm that the downloaded image is 1024*1024 and error if not.
• There are some unused imports and functions. These don't have to be dealt with in this PR, but can be identified with:

$ pylint pyveg/src/* | grep W0611

• (Plan B) If we still get lots of downloaded images that are not 1024*1024, then we could try using rasterio clip to the downloaded images to real-world coordinates, rather than pixel coordinates.

I've added a fail and logging error if the downloaded images do not have the right size or if the bounds are not the same as the input. It seems to be ok for the cases I have tested.

Also, I've checked that the chips created in the pipeline map to the geoparquet file (using chip_id) and that seems to be working well too :-)