Eomaps

</details>

☁️ minor (non-breaking) changes
- the default `radius_crs` for the `mark` callback is now determined based on the `radius_crs` assigned in the plot-shape
... this definition allows using `m.cb.pick.attach.mark(buffer=3)` directly without having to worry about the crs
(previously `in`, e.g. the input-crs was used by default)
- the background patch of the compass is now by default set to `None`

🔨 fixes
- fix issues with manual color specifications for various plot-shapes
(e.g. when providing explicit color-arrays via `m.plot_map(color=[...])` )
- fix issues with shapes close to crs-bounds
- support estimation of different x- and y- radius for 2D datasets
- warn if datapoints are masked or if datapoints are outside the CRS-bounds
- cache shape transformers (so they are not re-initialized all the time)

new
there are 2 new plot-types:
`"delauney_triangulation"`
- ... e.g. plot a (continuous) triangulation-shading of irregularly distributed points
- it supports additional customization by adding a suffix to the "shape"-name:
- `_flat`: plot actual polygons (with mean-values) instead of a mesh that interpolates values
- `_masked`: use the defined "radius" to mask any triangle for which the distance between the centroid and the vertices exceeds (2 * radius)
(particularly useful for densely sampled data-points whose exterior is a a concave shape)
- `_flat_masked`: a combination of the above options

`"Voroni"`
- create a voroni-diagram
- similar as with the Delauney triangulation, any polygon whose defining data-point is farther than (2 * radius) away from any vertex of the Voroni-diagram is masked

fixes
- cache the background on first draw of the map so that callbacks can immediately trigger

new
- example notebook updated to show most of the implemented functionalities
- markers & annotations now support `permanent=True/False`
- there's now an option `m.set_plot_specs(histbins="bins")` to use the bins calculated
by the data-classification for the histogram! (...only possible if a classification is used!)
- you can now specify "layers" for dynamic plot components (e.g. markers, annotations etc)
- option "radius_crs" added to `m.cb.mark` to specify the radius of the marker in an arbitrary crs
new functions:
- `m.cb.clear_annotations()` and `m.cb.clear_markers()` to clear all annotations and markers
- `m.figure.set_colorbar_position()` to manually change the position of the colorbar & histogram
- `m.add_overlay_legend()` to quickly customize the legend for map-overlays
fixes
- "rectangle"-marker can now be properly "buffered"
- fix setting "vmin" and "vmax" in `plot_specs`
- major improvements on blitting implementation
- use explicit cleanup functions to clear markers & annotations separately
- set "double_click=False" as the default for callbacks
(to avoid the open issue that double-clicks are not recognized in `ipympl` for Jupyter-Notebooks )
- ensure histogram is always properly positioned on top of the colorbar
- fix rotation and size of markers
- allow multiple callbacks (with same button) for "marker" and "annotation"
- fix initialization of subgrids if a `GridSpec` is passed to `m.plot_map`
- always overwrite all `classify_specs` to avoid mixing kwargs

fixes
- fix 16 (e.g. install issues with VERSION file)
- fix issues when coordinates are provided as integer-arrays and radius is estimated as float
- execute callbacks only if point is identified

fixes
- there were some issues when specifying `"radius_crs"` in conjunction with `"cpos"` that are fixed now
- `"radius_crs"` can be one of `"in"`, `"out"` or any crs-specification recognized by `pyproj`
- `"cpos"` can be one of `"c"`, `"ul"`, `"ur"`, `"ll"`, `"lr"`
- fix typo in README: `xcoord="lon"` and `ycoord="lat"`
- use explicit "VERSION" file to specify module-version in one place

new
- `trimesh_rectangles` has been added as a possible plot-shape (e.g. `m.set_plot_specs(shape="trimesh_rectangles")`
(it uses a triangular mesh for plotting which is useful for contour-plots since pixel-boundaries are invisible)
- option to copy the data with `m.copy(copy_data=True)` has been added

fixes
- copy now performs a deep-copy of all properties
- make `geopandas` an optional import (basic functions do not require geopandas)
- add `pyepsg` as dependency
- use `xcoord` and `ycoord` as labels for annotations
- cleanup callbacks if figure is closed