Quadmesh with 3D (RGB) array

[rabernat]

Is your feature request related to a problem? Please describe.

I would like to resample RGB data (shape H, W, 3) on a quadmesh grid, using datashader. This works with Canvas.raster but not Canvas.quadmesh.

Describe the solution you'd like

Canvas.quadmesh accepts 3D data, like Canvas.raster.

Reproducer

import datashader as dsh
import numpy as np
import xarray as xr

data = np.random.random((1000, 1000, 3))
da = xr.DataArray(
    data,
    coords={"x": np.linspace(-1, 1, 1000), "y": np.linspace(-1, 1, 1000), "band": [0, 1, 2]},
    dims=("y", "x", "band"),
  
)

cvs = dsh.Canvas(
    plot_height=512,
    plot_width=512,
    x_range=(-1, 1),
    y_range=(-1, 1),
)

data = np.random.random((1000, 1000, 3))
da = xr.DataArray(
    data,
    coords={"x": np.linspace(-1, 1, 1000), "y": np.linspace(-1, 1, 1000), "band": [0, 1, 2]},
    dims=("y", "x", "band"),
    name="foo"
)

# works
cvs.raster(da.transpose(..., "y", "x"))

# fails
cvs.quadmesh(da.transpose(..., "y", "x"), x='x', y='y')
# -> ValueError: ('y', 'x') must be a permuted list of FrozenMappingWarningOnValuesAccess({'x': 1000, 'y': 1000, 'band': 3}), unless `...` is included