Consider using cell types to determine `k` for batch ARI k-means

From this comment, https://github.com/AlexsLemonade/sc-data-integration/pull/101#pullrequestreview-1077651347, we may wish to extend functionality in the `calculate_batch_ari()` function to allow for selecting the k-means `k` based on the number of intersecting cell types across batches.

Start by locating the calculate_batch_ari() function and read the discussion in pull request 101, especially the linked review comment. Clarify how intersecting cell types across batches should determine k, then define the expected behavior before assessing implementation and validation needs.

Written by the indexing model from the issue text.