Generally comparable perf for the "good" case where memchr doesn't
return any collisions (false matches on lower byte) but clearly faster
with collisions.
bytesobject.c, bytearrayobject.c and unicodeobject.c now define all
macros used by stringlib, to avoid using undefined macros.
Fix "gcc -Wundef" warnings.
The public API symbols being removed are:
_PyBytes_InsertThousandsGroupingLocale, _PyBytes_InsertThousandsGrouping, _Py_InitializeFromArgs, _Py_InitializeFromWideArgs, _PyFloat_Repr, _PyFloat_Digits,
_PyFloat_DigitsInit, PyFrame_ExtendStack, _PyAIterWrapper_Type, PyNullImporter_Type, PyCmpWrapper_Type, PySortWrapper_Type, PyNoArgsFunction.
* Decode thousands separator and decimal point using PyUnicode_DecodeLocale()
(from the locale encoding), instead of decoding them implicitly from latin1
* Remove _PyUnicode_InsertThousandsGroupingLocale(), it was not used
* Change _PyUnicode_InsertThousandsGrouping() API to return the maximum
character if unicode is NULL
* Replace MIN/MAX macros by Py_MIN/Py_MAX
* stringlib/undef.h undefines STRINGLIB_IS_UNICODE
* stringlib/localeutil.h only supports Unicode
