Issue111 5 code cleanups that do not solve any problems, but perhaps prevent undiscoverred ones.

[petersilva]

while working on #111, found many random problems that did not cause obvious issues... but they are still wrong.
If we don't fix them now, eventually they will provoke issues so might was well be pro-active:

• there were some situations where errno was not set properly returning an error status (operation failed.)
• in exit_cleanup_posts() we want to post files that were not closed... but not sockets, or pipes, or devices, or... so test if it is a REGULAR file, and only do further work on those.
• with valgrind, found a memory "leak" (harmless) ... but fixed to reduce qvetching.
• error message about config file not found was not listing the obvious, normal place to look... It did look there, but the message said it didn't.... fixed that.
• earlier patch made exit_clieanup skip stderr... to get more error messages during process exit. That is great when debugging process exit, but normally, processes don't need that, and they want to post their stderr... so ideally should post it. but it is often needed when debugging. so made SR_SHIMDEBUG=128 trigger that behaviour.

[andreleblanc11]

I went through the commits, and I think it looks OK.