Fix but render_shapes() when coordinate system "global" not present

[melonora]

closes: #447

In render shapes, when using datashader, all transforms were obtained instead of the transform to the coordinate system which is supposed to be rendered. This was a problem for the visium notebook when trying the following code snippet:

import spatialdata as sd
import spatialdata_plot  
import matplotlib.pyplot as plt

sdata = sd.read_zarr("C:\\Users\\w-mv\\PycharmProjects\\spatialdata-notebooks\\notebooks\\examples\\visium_hd.zarr")

plt.figure(figsize=(10, 10))
ax = plt.gca()

gene_name = "AA986860"
sdata.pl.render_shapes("Visium_HD_Mouse_Small_Intestine_square_016um", color=gene_name, method="datashader").pl.show(
    coordinate_systems="Visium_HD_Mouse_Small_Intestine", ax=ax
)

After fixing that, the transform was applied and again parsed by ShapesModel. This ensured that after that one can only have a shape with coordinate system global. This means that the following should always use coordinate system global:

plot_width, plot_height, x_ext, y_ext, factor = _get_extent_and_range_for_datashader_canvas(
            transformed_element, "global", ax, fig_params
        )

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 85.21%. Comparing base (022a65a) to head (9793987).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #446   +/-   ##
=======================================
  Coverage   85.21%   85.21%           
=======================================
  Files           8        8           
  Lines        1759     1759           
=======================================
  Hits         1499     1499           
  Misses        260      260           

Files with missing lines	Coverage Δ
src/spatialdata_plot/pl/render.py	96.89% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[LucaMarconato]

@melonora why "global" here? What if "global" is not used in the dataset that is being plotted?

[timtreis]

@melonora why was this merged? I think Lucas concern is valid here

[melonora]

Already talked with Luca (sorry tagging does not work on github app) and it is due to the recent change in which global is not required. This caused errors for datasets not having global. However, here the element first gets transformed and after transform the default cs in case nothing is passed on is global. Hence, second call always uses global.

[timtreis]

But shouldn't we preserve whatever name the user gave the cs? I think we reuse that as the default title for a CS when plotting

[melonora]

I will have another look when I am at my laptop again. Note that the 'coordinate_system' parameter itself is not being altered here. The point is that after the transform there is no other cs but global.

[timtreis]

At the very end of basic.py we're assigning these as titles here:

spatialdata-plot/src/spatialdata_plot/pl/basic.py

Line 995 in 7716d43

if title is None:

[melonora]

just as a last follow up, this is not an issue as the title is still correspondent to the coordinate system used. I would have to follow up but it could be that the coordinate system argument in the second call where now the value by default is "global" can be deprecated.

For now just as final proof of the visium hd example in the notebooks. The following code snippet before was failing:

import spatialdata as sd
import spatialdata_plot
import matplotlib.pyplot as plt

sdata = sd.read_zarr("visium_hd.zarr")

plt.figure(figsize=(10, 10))
ax = plt.gca()

gene_name = "AA986860"
sdata.pl.render_shapes("Visium_HD_Mouse_Small_Intestine_square_016um", color=gene_name, method="datashader").pl.show(
    coordinate_systems="Visium_HD_Mouse_Small_Intestine", ax=ax
)

However, this PR fixes the issue of having the wrong coordinate system as argument in the second call and gives you this plot: