Figure out where to store similarity measure information

[jaroslavh]

Ideas are:

• let it be an argument while calling the script
• let it be a function of a Cluster

[jaroslavh]

Now I lean to it being an argument of a Cluster instance, need to think it through.