When updating the CPEs, the CPE dictionary file is no longer split
into separate temporary files. Instead a new XML file iterator added to
gvm-libs is used to avoid creating the DOM of the whole file.
Also, the buffered insert statements are now also run when a size
threshold configurable via a setting is reached.

This removes the dependency on the xml_split script, avoids problems
with the temp split XML files being left behind in case of crashes and
gives more control over the memory usage of the CPEs update.

The error handling is improved by freeing the iterator on errors,
consistently jumping to the failure handling and handling cases
where (re-)initalizing the parser context fails.
Also the limit for the threshold is now always applied.

This optional dependency is no longer needed with the new XML file
iterator.