Update `data_as_image` to return masked values

[JackDunnNZ]

Prior to this change, the raw data array was being returned, which could include nodata values, throwing off an automatic image color scale. In the following example, the nodata value used is -9999 (for all oceans), which causes the image to basically appear as a land mask:

from matplotlib.pyplot import imshow
from rio_tiler.io.rasterio import Reader

with Reader("https://data.chc.ucsb.edu/products/CHIRPS-2.0/global_daily/cogs/p05/2023/chirps-v2.0.2023.02.01.cog") as r:
    img = r.read(nodata=-9999)

imshow(img.data_as_image())

If instead we use the masked array values, the automatic color scale is scoped to just the valid data values:

imshow(img.array[0, :, :])

[vincentsarago]

@JackDunnNZ will imshow(img.data_as_image()) still work event if the return type is a Numpy Masked array?

[JackDunnNZ]

Yes it seems to, the second example above is passing a masked array to imshow. I also get the same result running imshow(img.data_as_image()) on this branch

[vincentsarago]

@JackDunnNZ can you add something in the changelog 🙏

[vincentsarago]

🙏