REGRESSION: Regressions introduced by rpcrt4 wine v2.9 sync r74904

Since r74904 where rpcrt4 was synchronized to latest Wine staging 2.9, we introduced several regressions.
See the test results here: https://www.reactos.org/testman/compare.php?ids=53335,53337,53338
The problems are:

• We lost my fix for CORE-6561 and associated hacks (this is partly my fault here because the changes were not guarded by _REACTOS_ defines, BUT! they were present in our local rpcrt4_ros.diff !)
• The last errors returned by cross-rpcrt4 function calls are all wrong / not correctly set, and this forced ekohl to introduce ugly advapi32 hacks in r75044, r75052 and r75053 (I think he wasn't aware of the rpcrt4 regressions when they happened).