This fixes compatibility with Poppler 22.03, which changes the types of
those arguments.

This doesn't yet extract the full level of detail of the location names,
that needs column separation logic beyond a regular expression. To properly
use the "flights" operated by SNCF we'd also need a IATA code based train
station lookup table eventually I guess.

Air France actually puts raster image Aztec codes in PDFs at he smallest
possible resolution (which makes perfect sense, just nobody else seems to
do that), which then is only a 27x27 pixel image.

The barcode decoder is very picky in pure mode if element sizes don't
match exactly.

Rendering a stroke on the edge would otherwise be cut off potentially.

That is, we skip any path containing curves or diagonal lines.

We have ki18n_version.h in (semi) public headers, but were missing its
include search path.

This fixes nearby cities being considered the same due to just considering
their coordinates. This showed for example in form of weird mis-groupings
in Itinerary.

For some reason a newer sample has 18.15 instead of 18:15 in the time
section.

Also with the newer sample the cancellation url isn't matched. I can't
be bothered to properly fix it right now, so at least not fail in that
case.

(cherry picked from commit dcc24e2f)

Helps e.g. Itinerary to pick the right service for retrieving realtime
data.

Also, retrieve full station names and arrival times from the PDF.

GIT_SILENT

BER::Element is exported meanwhile, so we don't have to built it into
the VDV certificate download again.

This fixes merge issues in KMail for some PDF documents without timestamps
and triggering two or more extractors.