Bitarray

-------------------
* add optional index parameter to `.index()` to invert single bit
* fix `sys.getsizeof(bitarray)` by adding `.__sizeof__()`, see issue 100

-------------------
* add PyType_Ready usage, issue 66
* speedup search() for bitarrays with length 1 in sparse bitarrays,
see issue 67
* add tests

-------------------
* support signed integers in `util.ba2int()` and `util.int2ba()`,
see issue 85
* deprecate `.length()` in favor of `len()`

-------------------
* Use `Py_ssize_t` for bitarray index. This means that on 32bit
systems, the maximum number of elements in a bitarray is 2 GBits.
We used to have a special 64bit index type for all architectures, but
this prevented us from using Python's sequence, mapping and number
methods, and made those method lookups slow.
* speedup slice operations when step size = 1 (if alignment allows
copying whole bytes)
* Require equal endianness for operations: `&`, `|`, `^`, `&=`, `|=`, `^=`.
This should have always been the case but was overlooked in the past.
* raise TypeError when trying to create bitarray from boolean
* This will be last release to still support Python 2.6 (which was retired
in 2013). We do NOT plan to stop support for Python 2.7 anytime soon.

-------------------
* add more tests
* C-level:
- simplify pack/unpack code
- fix memory leak in `~` operation (bitarray_cpinvert)

-------------------
* add official Python 3.9 support
* improve many docstrings
* add DeprecationWarning for `bitdiff()`
* add DeprecationWarning when trying to extend bitarrays
from bytes on Python 3 (`bitarray(b'011')` and `.extend(b'110')`)
* C-level:
- Rewrote `.fromfile()` and `.tofile()` implementation,
such that now the same code is used for Python 2 and 3.
The new implementation is more memory efficient on
Python 3.
- use `memcmp()` in `richcompare()` to shortcut EQ/NE, when
comparing two very large bitarrays for equality the
speedup can easily be 100x
- simplify how unpacking is handled
* add more tests