Feature/graph analysis merging

[Croydon-Brixton]

As discussed, I'm opening this PR to the work in progress graph-analysis-merging branch to speed up the PR process.

I will add more documentation and an overview of this PR once it's fully ready 😉

[Croydon-Brixton]

@herbiebradley @rdnfn Branch is now ready for review. I'm aware that a detailed review today will likely not be possible, so don't feel pressured to. For the publicly deployed branch we can do that after the report deadline if needed.

Main new features:

• NodeMap: An object to store node identifications between two GeoGraph objects. It's essentially a bipartite graph. This forms the basis for timestacking. It can be inverted easily.
• GeoGraphTimeline: The main new addition. Allows to stack several geographs in time, perform node identification between them (via .timestack()) and to analyse time-series of landscape, habitat, class and patch level metrics. The node identification also allows us to determine absolute and relative growth rates for nodes as well as classifying the node-dynamics type. While the functions for node-dynamics calculations, etc. are currently written in pure python, in my experiements they always exectued within under a second on graphs with ~2000 nodes. The major performance bottleneck are still edge computations when building up habitats or for getting e.g. avg_component_isolation across time. I think several of these might be alleviated sligthly by casting the performance relevant code into cython or numba-izing what can be transformed with numba. But that's definitely something for after the report 😉
• Connectivity=8 now supported: We can fix the GDAL problem via our node identification on a connectivity=4 polygonised raster image. Connectivity=8 (diagonal patch connectivity) is important becauses it is the usual definition in rasterised landscape ecology. So using connectivity=8 is required to compare our metrics to Fragstats/pylandstats. Connectivity=4 is still supported as before.

Other changes:

• TimedGeoGraph: A GeoGraph with a self.time attribute. Inherits all else from GeoGraph.
• Handy wrappers for adding and removing nodes from GeoGraphs
• Minor bugfixes (e.g. crs handling of GeoGraph, ...)
• Demo notebooks that show how to plot and perform the time-series analysis. I will still clean those up a bit today.

[Croydon-Brixton]

All done! Looks good to merge, I just made a commit adding a docstring to _add_node in geograph.py, and improving some type hints.

Thanks a lot, that's excellent!