Bitarray

-------------------
* add data descriptions: `.nbytes`, `.padbits`, `.readonly`
* allow optional `endian` argument to be `None` when creating bitarrays
* fix type annotation for `canonical_decode()`, 178
* frozenbitarray's padbits are now guaranteed to be zero
* add tests

-------------------
* optimize `.reverse()`, see 177
* allow negative (byte) indices in `.bytereverse()`

-------------------
* add calculating of canonical Huffman codes `util.canonical_huffman()`
and decoding thereof `util.canonical_decode()`, see 173
* allow creating "Huffman codes" from frequency maps with a single symbol
in `util.huffman_code()` and `util.canonical_huffman()`, see 172
* allow bytes-like argument in `.frombytes()` and `.pack()` - previously,
the arguments were limited to the `bytes` object, see 174
* allow bytes-like arguments in `util.deserialize()`
* add official [pyodide](https://pyodide.org/) support
* add [DEFLATE decompression](../examples/puff/) example
* optimize `.bytereverse()`
* optimize `delslice()` for cases like `del a[1:17:2]` when `a` is large
* fix `examples/huffman/compress.py` to handle files with 0 or 1 characters,
see also 172
* add `skipIF` decorator for skipping tests
* add tests

-------------------
* improve `resize()`, see 167
* optimize `copy_n()` by avoiding loops, 171
* `.bytereverse()` no longer sets unused pad bits to zero

-------------------
* enable building wheels for multiple platforms and Python versions using
pypa/cibuildwheel, see 165 and 170 (thanks Brian Thorne, hardbyte)
* use setuptools instead of distutils in `setup.py`, 168
* add missing type hinting for `.count()` step argument

-------------------
* add optional step argument to `.count()` method, see 162 and 163
* add tests