datannurpy v0.22.0

• change: export_db / export_app omit columns that are entirely null/empty from the exported JSON files
• add: public copy_assets() helper and copy_assets / base_dir parameters on export_db() / export_app()
• add: license fields on folders and datasets, with matching kwargs and export support
• change: scan APIs now use shared EntityMetadata via metadata= in Python and YAML
• add: warning when a remote scan exceeds the materialization cap (suggests sample_size)
• fix: folder discovery skips unreadable subdirectories (e.g. SFTP ACLs) instead of aborting
• fix: add_folder(depth="dataset") logs each dataset like other depth modes
• fix: time series detection — recognizes compact YYYYMMDD, handles 4-digit years sharing a constant 2-digit token, and splits mixed yearly/quarterly/monthly/daily granularities
• fix: include/exclude as a bare string no longer iterates character-by-character
• perf: depth: value materializes the source once (≤1M rows) and shares it across autotag, frequency, and pattern passes
• perf: frequency value counts via PyArrow (~27× on wide datasets); cleaner float/timestamp formatting
• perf: pattern frequency fully vectorized in PyArrow (~1.2×)
• perf: incremental scan batched — _seen updates, mark_datasets_seen, _match_path index, finalize cascade, orphan-tag detection, and last_update_date from cached mtime
• perf: add_database re-applies cached metadata for unchanged tables in one batch
• perf: enumeration assignment from frequencies looks up columns in O(1)