Scib-metrics

Added

- Add `progress_bar` argument to {class}`scib_metrics.benchmark.Benchmarker` {pr}`152`.
- Add ability of {class}`scib_metrics.benchmark.Benchmarker` plotting code to handle missing sets of metrics {pr}`181`.
- Add random score in case of aggregate metrics not selected to be used in scib autotune in scvi-tools, {pr}`188`.

Changed

- Changed Leiden clustering now has a seed argument for reproducibility {pr}`173`.
- Changed passing `None` to `bio_conservation_metrics` or `batch_correction_metrics` in {class}`scib_metrics.benchmark.Benchmarker` now implies to skip this set of metrics {pr}`181`.

Fixed

- Fix neighbors connectivities in test to use new scanpy fn {pr}`170`.
- Fix Kmeans test {pr}`172`.
- Fix deprecation and future warnings {pr}`171`.
- Fix lisi return type and docstring {pr}`182`.

Changed

- Replace removed {class}`jax.random.KeyArray` with {class}`jax.Array` {pr}`135`.

Changed

- Refactor all relevant metrics to use `NeighborsResults` as input instead of sparse
distance/connectivity matrices {pr}`129`.

Fixed

- Fix KMeans. All previous versions had a bug with KMeans and ARI/NMI metrics are not reliable
with this clustering {pr}`115`.

Added

- Update isolated labels to use newest scib methodology {pr}`108`.

Fixed

- Fix jax one-hot error {pr}`107`.

Removed

- Drop Python 3.8 {pr}`107`.