Division of blocks across the SODM beneficiary condition is wrong

[shippy]

The complex on-the-fly trial and block generation code in SODM does generate trials separately for each condition, but after they are divided into blocks, "friend" condition gets assigned block 1 from "friend" and "self" each, and "self" condition gets block 2. This results in a nonuniform distribution of trial repetitions, making analysis harder and makes inferences weaker.

This requires the following fixes:

• Load trial properties from CSV (pending Hybrid (both static and dynamic) trial generation #112)
• Rewrite the "SODM from MDM" article's section on counterbalancing
• Find a way to unit test this