improve performance for queries on sparse multipolygons

Currently, when querying large, but very sparse multipolygons as areas-of-interest, it can happen that a lot of CPU time is spent inside the _bbox <-> polygon_ testing, leading to poor performance:


In these situations, the current implementation of the `fip` package are not ideal (it assumes somewhat contiguous and not too unisotropic input polygons to work really well).

see also https://github.com/GIScience/oshdb/blob/0.5.9/oshdb-util/src/main/java/org/heigit/bigspatialdata/oshdb/util/geometry/fip/FastInPolygon.java#L72

//cc @SlowMo24