Psifr

This version changes the output of all analysis functions to improve compatibility, reorganizes low-level functions to create a new stats API, and adds a new Contributor Guide to the documentation.

* All analysis functions now return all information in columns instead of indexes. This change was made to allow support for calling Psifr functions from R using Reticulate and the {psifrr} package. It should generally improve compatibility with other table-based analysis packages, which generally do not support MultiIndexes. This change doesn't break any known code, but users taking advantage of the Index in analysis output will have to alter their code to set the Index after running the analysis.
* Low-level functions that were spread across multiple modules are now placed together in a new `psifr.stats` module. This module provides a unified low-level API for calculating statistics based on list-format data.
* All masker functions are now placed in a new `psifr.maskers` module.
* Unit tests have been reorganized to group together statistics of the same type (e.g., lag-, category-, or distance-based) to make it easier to share fixtures between tests.
* The build system has been updated to use a `pyproject.toml` file instead of `setup.cfg`.
* A new Contributor Guide has been added to the documentation to provide an overview of the organization of the codebase and provide detailed instructions for developing new analyses.

This version fixes support for Seaborn 0.12.0.

This version changes the build temporarily to avoid using Seaborn 0.12.0, which currently breaks some plotting functions.

This version adds analyses to examine transition statistics based on items recalled before the just-recalled item.

* Added `fr.lag_crp_compound` to measure conditional response probability based on the lag of both the previous and current transition.
* Added `fr.distance_rank_shifted` to test whether transitions reflect distance to items recalled in earlier output positions.
* A new tool, `transitions.sequences_masker`, to help with analyses examining sequences of multiple transitions.

This version adds statistics to measure the extent of category clustering in recall sequences.

* Added `fr.category_clustering`, which calculates list-based clustering (LBC) and adjusted ratio of clustering (ARC) for free-recall data.
* Added unit tests for LBC and ARC based on published papers.

This version adds support for analyzing prior-list intrusions. It also adds tools for working with semantic distance analyses. The documentation is expanded with examples for most functions.

* Marking of prior-list intrusions and analysis of the list lag of prior-list intrusions.
* Documentation of semantic distance analyses in the user guide.
* A utility for generating datasets from study and recall lists.
* Tools for working with semantic distance measures.
* Examples to illustrate most functions.
* Style reformatting using Black.
* A change to the automated testing system to use GitHub Actions.