Cwb-ccc

- add functionality for breakdown of anchors / slots

new features
- `cqp_unescape()` in `utils.py`
- `SubCorpora()` in `cwb.py`
- option to not remove nodes from context table in collocation analyses (`rm_nodes=True`)
- option to calculate unigram breakdown (`split=False`)
- option to disable cache invalidation (`inval_cache=True`)

deprecated features
- deprecate `discoursemes.py`

under the hood
- deal with overlapping names of filter and highlighting discoursemes
- use `trieregex` to build minimal RegExps for disjunction queries
- speed up `Undump()`
- ensure library directory exists
- avoid `fillna()` with downcast

dependencies
- drop python3.7 support
- add support for python3.11 and python3.12
- compile with Cython 3.0.0
- update to association-measures 0.3.0
- update to pandas 2.0
- ensure compatibility with [cwb-cads](https://github.com/ausgerechnet/cwb-cads) v0.3.0

further house-keeping
- installation of CWB from [debian package](https://sourceforge.net/projects/cwb/files/cwb/cwb-3.5/deb/)
- use venv rather pipenv for dependency management
- improve ccc-utils (include gespeR)

bug fixes and performance improvements for MMDA

interface changes
- introduce `SubCorpus` (instead of `Dump`) (closes 64)
- return single dataframe for `Dumps.collocates()` and `.keywords()` (closes 65)
- paths to directories now end on `_dir` instead of `_path` (`registry_dir`, `lib_dir,` `data_dir`)
- `Corpus._..._available` → `Corpus.available_...()` (`attributes`, `wordlists`, `macros`)

caching
- speed up collocation and keyword retrieval by caching interim results

maintenance
- update dependencies (do not use `pandas v2.x` for now)
- extend doc strings
- remove Pipfile.lock from repository
- add .flake8 file for ignoring long lines
- add a test showing correct functionality of `keywords()` when using several p-atts (closes 73)

ccc-utils
- remove unsupported packages (R 4.3.1)
- improve usability of concordancing function

- support for CWB 3.6.x (and above)

- bugfix: ignore unavailable corpora defined in registry when listing corpora to get equally sized lists of names and sizes