In the bug file, some text wrapping was badly computed when showing the
document (but not at the first rendering) due to that weird NotToWrap
mecahnism in SwLayouter. Checked the original issue i#40155 that
removing this does hurt.

This reverts commit e2a8fb04. This is a
feature, not a bug and would cause loads of documents to be badly
rendered.

Link to the spec that brough that feature:
    http://www.openoffice.org/specs/writer/compatibility/adjust-text-wrapping.sxw

Conflicts:

	sw/source/core/text/txtfly.cxx

This reverts commit 0589dd3f.

Given the find above, this assert will never be triggered anyway

This reverts commit 4762788e.

This is a ridiculous over-engineered thread-safe static string for
something which is called so few times, and costs 1k of long-term
memory as the OUString dtor needs to be put into a callback
queue to get called at module unload time.

Just return a new string every time, *rolls eyes*

There were multiple issues here:

- convertEmuToHmm() not handling negative values
- position:absolute style property being ignored
- mso-position-vertical-relative is not converted to
  text::RelOrientation
- SwAnchoredDrawObject::_SetPositioningAttr() re-positioning already
  positioned objects
- DomainMapper_Impl::PushShapeContext() inserting positioned objects as
  character

...otherwise bridges/source/cpp_uno/mingw_intel C++ ABI assumptions would not
meet what i686-w64-mingw32-g++ emits (though it is unclear to me how this shall
ever have worked before, as kendy claims it once did; also, at least for me,
"wine soffice.exe" still does not work, as there is apparently confusion about
C++ exception stack unwinding between GCC and cpp_uno/mingw_intel code).

db3786ee caused chart2_unoapi test to crash