Add all points of the edges to the path.

Add unit tests.

Don't limit the distance when finding the graph entry
and exit points, start/goal can be arbitrarily far away
from the graph.

Always return PathfinderResult, also in private
functions.

If the graph entry (closest vertex to start location) is closer
to the graph exit (closest vertex to the goal location) than the
range, the GraphPathfinder can't generate a useful path.

Return goalNodeInvalid == true in this case and log an error.

An edge with a single vertex results in an
infinite loop.

Once calculated for a grid cell, cache the penalty
in the pathfinder. The penalty depends currently only
on the position and not the heading, so this requires
a limited amount of memory.

This should reduce the calls for getNodePenalty() by
an order of magnitude, but needs to be tested on a
bigger field with difficult situations.

Also log should now indicate which pathfinder
is being used.