v0.0.2

Added

• ds.read_imaginary_metrics: samples one row per unique benchmark configuration from the metrics table and relabels it as 'ImaginaryMethod', useful for baseline comparisons or testing visualizations.
• get_terms: retrieves the filtering terms for a given dataset as a dictionary mapping evaluation database names to lists of terms.
• tl.cre_to_tss_distance: computes the distance from each CRE in a GRN to the promoter window of its target gene, using the Promoters database. Accepts a single GRN DataFrame or a dictionary of GRNs.
• pl.cre_to_tss_distance: plots CRE-to-TSS distance distributions as horizontal boxplots, with an optional vertical threshold line (default 250,000 bp).
• Tutorial notebooks.
• Added scanpy and xgboost as explicit dependencies; removed the [full] extra from decoupler.

Changed

• Dataset filtering terms moved from inline Python dicts in config.py into bundled YAML files (src/gretapy/data/), loaded at import time. This reduces config.py from ~2,600 lines to ~260 lines with no change to the public API.
• pl.links: legend is now displayed once on the middle TF panel (with multi-column layout for many GRNs) instead of being repeated on every subplot.
• mt.correlation now filters for promoters first and then computes correlations, reducing the number of compute.