Harmonize artifical dataset creation

Is your feature request related to a problem? Please describe.
There are multiple artificial dataset creation functions. It should be clear which ones are most useful and when.

Describe the solution you'd like
Merge or document the different artificial dataset implementations. Ideally, the default one and the benchmarking one are merged and the ones from libraries using SpatialData can reuse some functionality to make more specific artificial datasets.

Additional context
Here is a list of some implementations:

• spatialdata.datasets.blobs
  • default basic option, slow and limited in use
  • https://github.com/scverse/spatialdata/blob/main/src/spatialdata/datasets.py
• from benchmarks.utils import make_blobs
  • https://github.com/berombau/spatialdata/blob/benchmark-asv/benchmarks/utils.py
  • very fast using adapted code from https://github.com/napari/napari/blob/195bbd0720fce1bae665cd18ccee5456a095b830/napari/benchmarks/utils.py#L175
• SOPA blobs
  • https://github.com/gustaveroussy/sopa/blob/f1f5a99ee7f5a9489e511241a3a62bb520ec9860/sopa/utils/data.py#L188
  • more irregular cell shapes, genes from list
• Harpy cluster_blobs
  • https://github.com/saeyslab/harpy/blob/main/src/sparrow/datasets/cluster_blobs.py
  • multisample, multichannel, ground truth cell type annotation

Start by comparing spatialdata.datasets.blobs in src/spatialdata/datasets.py with benchmarks/utils.py and the linked SOPA and Harpy implementations. Identify the overlapping capabilities and differences, then clarify whether the expected outcome is a shared dataset API, merged implementations, documentation, or some combination; completion criteria are not defined in the issue.