[skip ci]

* Fix coverage story and add coveralls.io.

* Always be verbose to avoid failing builds because of 10 minutes without output.

* Add badges.

Drop Python 3.3 and add Python 3.7. 

Use operator.index to avoid DeprecationWarning on Python 2

Fixes #79

Slightly faster on CPython:

$ python -m perf timeit -s 'from struct import Struct; s = Struct("i"); unpack=s.unpack; pack=s.pack' 'unpack(pack(123456))'
.....................
Mean +- std dev: 252 ns +- 12 ns
$ python -m perf timeit -s 'from struct import Struct; s = Struct("i"); from operator import index; pack=s.pack' 'pack(index(123456))'
.....................
Mean +- std dev: 235 ns +- 6 ns

Very slightly faster on PyPy:

python -m perf timeit -s 'from struct import Struct; s = Struct("i"); unpack=s.unpack; pack=s.pack' 'unpack(pack(123456))'
.........
Mean +- std dev: 4.24 ns +- 0.12 ns
$ python -m perf timeit -s 'from struct import Struct; s = Struct("i"); from operator import index; pack=s.pack' 'pack(index(123456))'
.........
Mean +- std dev: 4.19 ns +- 0.10 ns

Support PURE_PYTHON at runtime. 

Detect pip installs with 'bdist_wheel' (or manual builds for upload) and don't let the C extensions fail to build unless PURE_PYTHON is set.

This makes the test runs in pure_python mode 15s faster.

Microbenchmarks on CPython 2.7.14:

$ python -m perf timeit -s 'from struct import pack,unpack' 'unpack("i", pack("i", 123456))'
.....................
Mean +- std dev: 365 ns +- 12 ns

$ python -m perf timeit -s 'from struct import Struct; s = Struct("i"); unpack=s.unpack; pack=s.pack' 'unpack(pack(123456))'
.....................
Mean +- std dev: 251 ns +- 9 ns

No appreciable difference on PyPy.

Add failing tests and then fix them. Use a common function to DRY in
all the modules.

Fixes #78

Also use zope.testrunner in tox (and travis) because it allows passing
arguments to select what tests to run, see slow tests, etc.

Simplify setup.py to always build the C modules and avoid polluting
wheel caches (the same as zope.security and other projects that have
been worked on lately).

Bump the minimum 'persistent' requirement to one that actually works
in PURE_PYTHON mode (plus fix wheel metadata; otherwise we need
environment markers).

Test PURE_PYTHON in environments on Travis. Do some minor tuning to
eliminate redundant test cases and speed up test time.

Fix possibility of rare crash during dealloc. Fixes #75

cleanup after manylinux wheelbuilding

errant white space preventing some builds in manylinux

Build and deploy manylinux builds on tagged commit

Add documentation for Length class

Conflicts:
	CHANGES.rst
	setup.py

I tried to tighten the whitelist, but that didn't work.

distutils must die.

Work around AttributeError raised by Python 3.5.x.

Alternative to PR #55.

* Use macpython everywhere.

Add 3.6 for mac.

See zopefoundation/persistent#59

* Manually install persistent for travis.

* plain pypy3 is python 3.2 which won't run newer setuptools

Add support for Python 3.6

Allow None as a special key (sorted smaller than all others).

Cuz the complicated version was wrong, in addition to being complicated :)