Bug in many body kernel

[nw13slx]

For the many body kernel, the list of distances (and their vectors) is stored in the bond_array_mb attribute in the AtomicEnvironment object. But the corresponding atomic species are not stored.

In the kernel/mc_simple.py, the kernel uses the env.etypes to determine the atomic species, assuming that the order of atom showing up in bond_array_2 and bond_array_mb is the same (only the later one being a shorter list).

Previous unit tests passed, simply because of luck... in those tests, the many body cutoff and two-body cutoff were the same. The bond_array_mb and bond_array_2 are exactly the same. That's why the test did not diagnose this bug.

This bug is fixed in 2b8e90d , but also with more functions to using different cutoffs for different interactions.

I'll only a new branch solely for this bug...