Support for different heuristics for "the best triangle" to be included to a generated meshlet

[JarkkoPFC]

Currently the meshlets are created greedily, by expanding the generated meshlet by adding the next "best triangle" adjacent to the generated meshlet. "The best triangle" is currently defined based on the triangle distance from the generated meshlet center, which results in "scale look" and optimizes the bounding sphere of the meshlet. However, it could be useful to have other heuristics to define the "best triangle", e.g. using the similarity of the triangle normals instead to favor meshlet "flatness".