What is the optimal orientation to place 4 reaction wheels, such that complete 3 axis control can be maintained at all times?
Most often they're configured in a "pyramid", which is done for both controllability and redundancy reasons (it's tolerant to one wheel failure).
That said, there are many subtleties to this problem, and all spacecraft with reaction wheels do not necessarily follow the same design. The paper "Sun-Tracking Commands and Reaction Wheel Sizing with Configuration Optimization" by Hari Hablani is available here [PDF], and discusses reaction wheel sizing and configuration considerations.