Clevercsv

* Updates to release process. This version introduces pre-compiled wheels for
Python 3.9.

* Add an `encoding` argument to `write_table` to allow specifying the output
encoding. Thanks to mitchgrogg for reporting [issue
27](https://github.com/alan-turing-institute/CleverCSV/issues/27).

* Add support for standardizing in-place and standardizing multiple files.
* Add warning on duplicate field names in DictReader
* Add return value to writers to match the standard library.

* Various speed ups to constructing the list of potential dialects. This
removes a costly step of the detection process that will likely add a few
more potential dialects, but has the end result of making overall dialect
detection faster.

* Rename wrapper functions to a more coherent naming scheme. Old names will be
available until 0.7.0, but now produce a FutureWarning.
* Add ``stream_dicts`` wrapper function.
* Improve handling of file encoding for the ``read_dataframe`` wrapper:
detected encoding is now passed on to Pandas.
* Fix handling of optional dependency error for TabView on non-Windows
platforms.

* Update URL regex to avoid catastrophic backtracking and increase
performance. See [issue
13](https://github.com/alan-turing-institute/CleverCSV/issues/13) and
[issue 15](https://github.com/alan-turing-institute/CleverCSV/issues/15).
Thanks to kaskawu for the fix and jlumbroso for re-raising the issue.
* Add ``num_chars`` keyword argument to ``read_as_dicts`` and ``csv2df``
wrappers.
* Improve documentation w.r.t. handling large files. Thanks to jlumbroso for
raising this issue.