Eomaps

A minor bugfix release.

🌳 new
- there is now an option to completely disable extension arrows when adding a colorbar
(e.g. with `m.add_colorbar(add_extend_arrows=False)`)

- it is now possible to set the "layer" of an inset-map on initialization (e.g. `m.new_inset_map(layer="adsf")`
(by default, the `"all"` layer is used for inset-maps!)

🔨 fixes
- fix using alpha-transparency with wms and wmts services (e.g. `m.add_wms.<service>.add_layer.<layer>(alpha=0.5)`)
- fix colors of histogram-bins if the bins extend over more than one colorbar-split
- fix position of colorbar extension-arrows after using `m.subplots_adjust`
- fix appearance of "vertical" colorbar extension arrows

A minor bugfix release.

🌳 new
- To address issues when plotting **2D raster-data** with **unsorted coordinates**, there is now an option
to force sorting the data by coordinates prior to plotting with `m.plot_map(assume_sorted=False)`
- only works if coordinates are provided as 1D arrays and the data is provided as a 2D array
- sorting is only relevant for `raster` and `shade_raster` shapes (other shapes don't require sorting)
- by default no sorting is performed!
🌦️ changes
- The pick-radius used to identify clicked pixels can now be set in 2 ways `m.plot_map(pick_distance=...)`:
- if a number is provided, it is used as a multiplier for the dataset-radius
e.g. `pick-radius = pick_distance * radius_in_plot_crs`)
- if a string is provided, it is directly used as the pick-radius in units of the plot_crs
e.g. `pick-radius = float(pick_distance)`
🔨 fixes
- ❗ fix using "pick" callbacks with `m.set_shape.raster()` and 2D input-coordinates
- make radius-estimation more robust (fallback to nearest-neighbor if 2D estimation fails)
- fallback to `np.inf` for pick_distance if radius estimation fails
- fix delaunay_triangulation does not need to estimate the radius if masked=False
- fix issues with registering pandas

A bugfix release that greatly reduces import time!

🌦️ changes
- 🕐 greatly reduced "plain" import-time
- "heavy" modules such as 'datashader', 'xarray' etc. are now lazily imported only when needed
- this reduces runtime of `from eomaps import Maps` in a fresh terminal from ~5sec to ~1sec
- 🍃 `m.add_marker(...)` now supports the "layer" kwarg
- 🍃 `m.show()` now automatically calls `plt.show()` to actually draw the figure if matplotlib is in non-interactive mode

🔨 fixes
- legends for WebMap services are now updated based on the visible layer
- `m.add_marker` now only draws newly added markers (instead of updating all dynamic artists)
- avoid using `pkg_resources` to identify package-version (use explicit `_version.py` instead)

🌳 New
- ⭐ there's a new function `m.fetch_layers()`
- ...to fetch (and cache) all layers of a map so that switching layers (with buttons & sliders) is fast
- a associated callback `m.cb.keypress.attach.fetch_layers()` is also available
- ⭐ there's a new pre-defined WebMap service for the "Copernicus Atmosphere Monitoring Service"
- accessible via `m.add_wms.CAMS` (see [ECMWF CAMS](https://confluence.ecmwf.int/display/CKB/WMS+for+CAMS+Global+and+European+air+quality+products) for details )
- 📃 docs on [WebMap services](https://eomaps.readthedocs.io/en/latest/api.html#webmap-layers) have been updated with examples on how to
- use custom service-urls
- pass arguments (time, style etc.) to the service

🌦️ Changes
- 🍃 `m.set_shape.raster()` can now be used with 1D `pandas.DataFrames` as well!
(DataFrames will be re-structured internally based on unique values of the coordinate-columns)
- 🍃 `.add_legend()` for WebMap services can now properly recognizes the used map-style
`m.add_wms.<SERVICE>.....add_layer.<LAYER>.add_legend()`

🔨 Fixes
- ❗ fix pixel centers shifted by half pixel-size with `m.set_shape.raster()`
- speedup `m.set_shape.raster()` (remove unnecessary calculations)
- fix identification of picked ID if `pandas.Series` objects are used as coordinates
- allow setting shade-aggregator via strings (e.g. "mean", "min", "max" etc.) for `m.shade_points`
- fix colorbar kwargs assertions should intercept before axes are created
- add fix for QuadMesh performance issues for older matplotlib versions (e.g. < v3.5.2)
- fix issues with updating WebMaps when using `m.util.layer_slider()`
- fix re-using WebMap objects with different kwargs
- fix fetching WebMap legends for different styles

A bugfix release that brings some nice updates for the colorbar.

🌈 updates for the colorbar
- ⭐ Colorbars now have proper extension-arrows to indicate values outside the colorbar-range
- ⭐ `m.add_colorbar()` supports some new arguments to further customize the colorbar:
- `histogram_size` : set the relative height of the histogram in relation to the height of the colorbar
(use 0 for a plain colorbar and 999 for a plain histogram)
- `extend_frac`: set the relative size of the extension-arrows
- `add_extend_arrows`: toggle drawing extension-arrows
(by default arrows are only drawn if there are values outside the [`vmin`, `vmax`] range)

🌦️ changes
- setting vmin/vmax no longer clips data-values
(e.g. callbacks now always retrieve actual data-values independent of vmin/vmax)

🔨 fixes
- fix setting zorder for WebMap layers
- make sure the default zorder for `m.plot_map()` is 1 for all shapes
- fix test-action issues (use mamba + updated version of conda-incubator)
- fix adding static annotations by dataset-ID with `m.add_annotation()`
- fix some issues with draggable axes
- fix sorting of callback-execution order

A minor release that brings an awesome new feature and some bugfixes.
🔬 New feature: Inset-maps!
- Quickly create zoomed-in views on selected regions of a map with `m.new_inset_map()`

Checkout the docs for more details: [🔬 Inset-maps - zoom-in on interesting areas](https://eomaps.readthedocs.io/en/latest/api.html#inset-maps-zoom-in-on-interesting-areas)



🔨 fixes
- allow using a scalar radius for calculating ellipse and rectangle points
- make sure the blit-manager always uses the correct figure object
- make histogram plot background visible (but set boundary color to none)
- fix wrong error-catching for layer-slider
- fix issues with colored histograms for discrete colormaps