The _getFileType already process decodestring() with the data.

All minor changes are fixes to follow up changes from external sources.
LibreOffice and Imagemagick in this case.

The correspondent reasons are appropriated documented in the code. Switch
some skips to expected failures.

It is required to use TestCase from backportUnittest in other to report
expected failures in an appropriated way.

Those OOo tests are making the current tests stalle and test ffmpeg will
fail due lack of ffmpeg itself.

Word mimetype outcome changed, after update of libreoffice.

After several updates of libreoffice the list had some changes due the
new version of libreoffice binary.

This initial implementation has several copied code from ERP5 Project which
must be removed as soon as possible. The code should be comes from
erp5.util library.

This reverts commit 1346047c.

By default it is 800px but as we normally use exported content in root ERP5 web sites smaller defaults is usually better (still it's an ugly hack).

Use encoded text convertion by default.
Do not use filter which rely on java.