Pylangacq

Added
* `Reader` objects can now be concatenated by the addition operator `+`.
* Implemented the `head`, `tail`, and `info` methods at `Reader`.
* Added support for Python 3.10.
* Turned on Windows testing on CircleCI.
* Added `pyproject.toml`. Related to prioritizing `setup.cfg` for specifying
build metadata and options.

Changed
* The `to_strs` and `to_chat` methods of a `Reader` object return
tabulated outputs by default.
* Prioritized `to_chat` for the single file output use case.
* Unzipping CHAT data now uses less memory.
* Switched to `setup.cfg` to fully specify build metadata and options,
while keeping a minimal `setup.py` for backward compatibility.
Related to the new `pyproject.toml`.
* Switched the Sphinx docs theme from `sphinx-rtd-theme` to `furo`.

Removed
* Dropped support for Python 3.6.

Security
* Turned on `safety` and `bandit` checks at CircleCI builds.

Added
* `Reader.from_zip` (also `read_chat`) now keeps the downloaded ZIP archive
in a non-temporary directory for possible re-use.
- Added the kwarg `use_cached` in `Reader.from_zip`, so that we use the cached data
by default for the same input URL, and that we can force re-downloading by
setting `use_cached` to `False`.
- Added the kwarg `session` in `Reader.from_zip`, in case using a customized
`requests.Session` instance is desired. `session` also makes it possible to
write tests for the new kwarg `use_cached`.
- Added the helper functions `cached_data_info` and `remove_cached_data`.
* `Reader` has the new `to_strs` method that yields CHAT data strings.
* `Reader` has the new `to_chat` method that exports data to local files.

Changed
* CHAT parsing for the header information is now more robust for varying whitespace
characters between the head and its associated value.

Removed
* Dropped kwarg `allow_remote` in `Reader.from_zip`. This kwarg wouldn't make any sense
anymore, or at least would be confusing with the introduction of `use_cached`.

Fixed
- The header/metadata has a more reasonable representation for emptiness
when input data is empty.

Added
* Added the `parallel` optional argument to the `Reader` methods
`{from_zip, from_dir, from_files, from_strs}`
so that parallelization can be turned off if desired.
* Added the `filter` method to `Reader` for filtering data by file paths.

Fixed
* The methods `append`, `append_left`, `extend`, and `extend_left` now work with a subclass
of `Reader`, not just `Reader` itself.

Fixed
* Fixed utterance cleaning so that it is now compatible with all CHILDES datasets.