Splink

- Dataframes to be registered when using `compare_two_records`, to avoid problems with data typing (because the input data can have an explicit schema) ([2493](https://github.com/moj-analytical-services/splink/pull/2493))

Added

- `cluster_pairwise_predictions_at_multiple_thresholds` to more efficiently cluster at multiple thresholds ([2437](https://github.com/moj-analytical-services/splink/pull/2437))

Fixed

- Fixed issue with `profile_columns` using latest Altair version ([2466](https://github.com/moj-analytical-services/splink/pull/2466))

Added

- Cluster without linker by RobinL in https://github.com/moj-analytical-services/splink/pull/2412
- Better autocomplete for dataframes by RobinL in https://github.com/moj-analytical-services/splink/pull/2434

Added

- Match weight and m and u probabilities charts now have improved tooltips ([2392](https://github.com/moj-analytical-services/splink/pull/2392))
- Added new `AbsoluteDifferenceLevel` comparison level for numerical columns ([2398](https://github.com/moj-analytical-services/splink/pull/2398))
- Added new `CosineSimilarityLevel` and `CosineSimilarityAtThresholds` for comparing array columns using cosine similarity ([2405](https://github.com/moj-analytical-services/splink/pull/2405))
- Added new `ArraySubsetLevel` for comparing array columns ([2416](https://github.com/moj-analytical-services/splink/pull/2416))

Fixed

- Fixed issue where `ColumnsReversedLevel` required equality on both columns ([2395](https://github.com/moj-analytical-services/splink/pull/2395))

Added

- When using DuckDB, you can now pass `duckdb.DuckDBPyRelation`s as input tables to the `Linker` ([2375](https://github.com/moj-analytical-services/splink/pull/2375))
- It's now possible to fix values for `m` and `u` probabilities in the settings such that they are not updated/changed during training. ([2379](https://github.com/moj-analytical-services/splink/pull/2379))
- All charts can now be returned as vega lite spec dictionaries ([2361](https://github.com/moj-analytical-services/splink/pull/2361))


Fixed

- Completeness chart now works correctly with indexed columns in spark ([2309](https://github.com/moj-analytical-services/splink/pull/2309))
- Completeness chart works even if you have a `source_dataset` column ([2323](https://github.com/moj-analytical-services/splink/pull/2323))
- `SQLiteAPI` can now be instantiated without error when opting not to register custom UDFs ([2342](https://github.com/moj-analytical-services/splink/pull/2342))
- Splink now runs properly when working in read-only filesystems ([2357](https://github.com/moj-analytical-services/splink/pull/2357))
- Infinite Bayes factor no longer causes SQL error in `Spark` ([2372](https://github.com/moj-analytical-services/splink/pull/2372))
- `splink_datasets` is now functional in read-only filesystems ([2378](https://github.com/moj-analytical-services/splink/pull/2378))