Squidpy dev (the-future)

Features

• squidpy.experimental.im.calculate_image_features() now featurizes tiles on a shared dask engine: n_jobs > 1 runs worker processes via a dask.distributed.LocalCluster (or an active Client), and per-tile BLAS/OpenMP threads are pinned to avoid oversubscription. This also speeds up the serial path. squidpy.experimental.tl.calculate_tiling_qc() shares the same engine. Adds distributed and threadpoolctl as dependencies.

• Fix squidpy.tl.var_by_distance() behaviour when providing numpy arrays of coordinates as anchor point.

• Update :attr:squidpy.pl.var_by_distance to show multiple variables on same plot. @LLehner #929