* better and faster trip clustering: trip tries
* add --write-colors to extract line colors from OSM data
* refactor config parameter names, update default pfaedle.cfg
* add --stats for writing a stats.json file
* add --no-fast-hops, --no-a-star, --no-trie for debugging
* general refactoring
* Enable transit line penalties also when no explicit line was given
Currently, if no transit line is given, all edges get penalty factor 0
regardless of whether there are transit lines following the edge or not.
The change introduced in this commit will give penalty factor 0 if there
is at least one transit line following the edge and penalty factor 1 if
there are no tranist lines at all following the edge.
Behavior in the case when a transit line is given is untouched.
This should improve routing results by preferring edges with
associated transit lines. This used to be the behaviour before
some refactoring. Maybe it was broken unintentionally.
If I'm not mistaken, there is a typo in the OsmBuilder when an edge is split
into two edges to insert a station: one of the new edges is registered with
the wrong geometry in the edge grid. This should fix it.
Changes:
* support for multiple GTFS feeds as input in filtering, read default global and local configuration files
* be a bit more memory conservative
* make caching optional
* dont delete orphan edge if it would transform a degree 3 node with a possible full turn into a degree 2 node eligible for contraction
* dedicated filters for funicular and gondola
* make max snap level option more intuitive
* allow filter rules for level 0
* additional fallback for station snapping
* dont try to route for MOT unequal to trip in -T mode, force-snap to orphaned OSM station if not snap was possible
* write bounds to filtered osm
* remove unused surrounding heuristic
* use bus lanes info
* be a bit more tolerant for bus oneway streets
* create missing directories
* error if no cfg is present, clean up evaluation Makefile
