Bots receive incorrect timesteps in Scenario

[jagapiou]

Due to an errant - bots were sent timesteps (1, ..., num_bots + 1) rather than (num_bots, ...) as intended. This was fixed in 09b69a5.

This will completely mess up the behavior of the bots on the scenarios as they don't see their own observations, and so any evaluation run using before this fix will be incorrect and should be rerun.

(NOTE: this didn't affect our experiments run for the Melting Pot paper since these weren't run using Scenario.)