Previously, if incrementing by the length failed to make the
index positive, this *incremented* index was passed in an attempt
to raise the correct error.

This fixes #1700.

Now uses specialised functions for complex abs

Print no message if force=True

Closes #1664

add a compile test for github issue #1692: raising StopAsyncIteration without using "await" was missing utility code and failed to compile

Closes #1692

use different codec alias names in test code than in optimiser code to check that the alias name resolution works

Repair incorrect C-API signature usage for PyUnicode_DecodeUTF16() when optimising bytes decoding from UTF-16(LE/BE)

Closes #1696

Remove obsolete documentation

Fix bad indentation in a pyximport test

The indentation was inadvertently broken when expanding tabs in
e908c0b9.

Fixes:

    $ python pyximport/test/test_reload.py
      File "pyximport/test/test_reload.py", line 23
        assert hello.x == 1
        ^
    IndentationError: unexpected indent

Fix incorrect use of flags in re.sub()

The 4th argument of re.sub() is maximum number of substitutions, not
flags.

re.M affects only semantics of ^ and $, so it wouldn't have any effect
on this regexp.

Update language_basics.rst

Allow "cdef inline" with default values in .pxd

Add declarations for several PyLong_Type functions

Add declaration for PyImport_ImportModuleLevel

Also doesn't require gil for int and long abs.

Notes:
* I don't think the correct utility code is loaded for complex abs
  (but it should be loaded by the complex data types hopefully)
* Would probably be worth making double and float abs not require gil

Fixed #1647

set.update() is actually variadic.  Fixes #1645.

Add create_extension() hook