fix setobject.c refcounts and error checks.

Fixed a few bugs on cjkcodecs:
- gbk and gb18030 codec now handle U+30FB KATAKANA MIDDLE DOT correctly.
- iso2022_jp_2 codec now encodes into G0 for KS X 1001, GB2312
  codepoints to conform the standard.
- iso2022_jp_3 and iso2022_jp_2004 codec can encode JIS X 0213:2
  codepoints now.

place of sys.stdout while running tests. Removing one more test
to make buildbot happy.

Let's try that once more. Buildbots were broken last time, but probably
because tests were sending data to stderr for testing it (sending to a
file doesn't touch the problem).

The fix is still the same, but tests were reduced (removing tests to
be able to fix something is weird, but oh well).

 (backport from rev. 51785)

Add asserts to check for 'impossible' NULL values, with comments.
In one place where I'm not 1000% sure about the non-NULL, raise
a RuntimeError for safety.

This should fix the klocwork issues that Neal sent me.  If so,
it should be applied to the release25-maint branch also.

Also make sure that every exception class has __module__ set to
'exceptions'.
 (backport)

imports (if not in package and if beyond toplevel package).
 (backport from rev. 51765)

of os.urandom().
 (backport from rev. 51762)

making subprocess not close standard file descriptors.

to use "in".
 (backport from rev. 51740)

Backport 51669: Make sure memory is properly cleaned up in file_init (even though this should not be able to happen since we already parsed the args)

generator expressions (x for x, in ... ) works again.

Sigh, I only fixed for loops the first time, not list comps and genexprs too.
I couldn't find any more unpacking cases where there is a similar bug lurking.

Partially from SF patch #1551339, but also taken from head.

checking to live happily with recent gcc optimizations that
assume signed integer arithmetic never overflows.

  Doc fix: hashlib objects don't always return a digest of 16 bytes.

 (backport from rev. 51677)

- various minor cleanups for improved consistency