(related to i#95158, but not quite)

this local declaration makes no sense, based on the previous code, it seems
this is a rest of the former code change

Show either a SPIN_DOWN arrow if there is a header or a plus is there is
none.

This fixes fdo#32861, by processing all remaining events after the
post-yield callback comes around (that's supposed to be the time
when the app can spend time on event processing).

  * this enables "make all dev-install" or "make all install" calls

Don't mark the template class TableDataHandler for DLL export/import
either. It seems to break the MSVC build, and dropping the decoration
fixed the problem.

Odd that I didn't notice before now, or is it really over a month
since I managed to build master last? That wouldn't be too surprising,
I guess. But other people do claim the have built master with MSVC
relatively recently. Or is it just that this templace class has only
recently been used more than earlier? Or does the fact that I am still
attempting a --enable-dbgutil build affect this? The world is full of
little mysteries.

Regression introduced with 4fae740d;
caused forms/qa/unoapi check to fail.

 + Implemented http://wiki.services.openoffice.org/wiki/DocumentBorder#Option_3
 + Show the text boundaries for the currently edited area, e.g. don't
   show the header and footer boundaries when editing the main body
 + Included the footnote boundaries into the main body ones as it's the
   same editing area
 + Reimplemented the painting of the manual page breaks. These are
   hidden when editing headers/footers and are be toggled with the
   non-printing characters instead of the document boundaries.

This reverts commits:
  * 7d700ecf
  * 523ff044
  * f7da7938
  * 28d1c250