I could not compare b_part.type() to the actual fluid particle types,
because those are not exposed in RNA_blender_cpp.h yet. We could probably
expose the missing types like PART_FLUID_FOAM, but I did not want to
investigate that yet, before getting some feedback on the solution.

@Sebastián Barschkis (sebbas), is my assumption that these particles don't have persistent ids
correct? Since the number of particles changes, those ids would be necessary
to find the same particle in different frames.