Support multiple levels of "acceptability" in PRIM.

[EwoutH]

Currently PRIM only supports a single threshold value, and an outcome is classified binary.

In some cases, it might be beneficial to actually input multiple levels of how good an outcome is, such as "unaccaptable", "accaptable" and "preferred". As side for support in Prim init, the Prim methods should also be able to support this:

• find_box() should be able to find boxes on multiple levels (with the level as input)
• with a selected box, inspect should show the multiple levels, with RdYlGn colormap when plotting.
• other Prim methods should also be adjusted.