Metrics should not break for different indices.

[mmschlk]

ATM, computing test metrics (MSE, MAE, SSE, ...) will break with a ValueError if two interaction indices are compared with different indices. While this might be good for technical reasons (some indices cannot semantically be compared to others) for an experimentation tool this might be too restrictive.

Proposal: In instances like described above, not a ValueError should be returned but a Warning and the metrics should be made compatible.