Trying to guess layout from flat ODF output is an approach that doesn't
scale, think of complex documents with split tables, etc.

Do it similar to the PDF export instead: take a metafile of each page
and use the existing SVG writer to embed it into XHTML.

Change-Id: I6e860834beb8025519d3e367f858077ae9e9c006
Reviewed-on: https://gerrit.libreoffice.org/45647Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Miklos Vajna <vmiklos@collabora.co.uk>