Discrepancy between paper and code in `pearson_delta`

[marvinsxtr]

Hey, thank you for this great evaluation package.

I noticed in the STATE paper on page 35, the Pearson Delta uses the absolute difference while the implementation uses the raw value (abs=False, see https://github.com/ArcInstitute/cell-eval/blob/main/src/cell_eval/metrics/_anndata.py#L211 and https://github.com/ArcInstitute/cell-eval/blob/main/src/cell_eval/metrics/_anndata.py#L31)

Is this intended? Maybe I missed something.