Add distributed primitives

Once #117 is merged, and #166 is realized, we'll want users to be able to use Dagger for traditional concurrent programming. In order to do so, we need to be able to ensure certain scheduling patterns and communication between thunks. Important primitives include:
- [ ] Barrier
- [ ] Co-scheduling (two or more thunks are guaranteed to be scheduled concurrently)
- [ ] Co-execution (same as co-scheduling, *and* the thunks will execute on the same type of processor (for shared-memory operations))
- [x] Wait on thunk completion (one or more)
- [x] Get thunk result
- [ ] Block/cancel thunk scheduling

The above primitives would enable https://github.com/JuliaParallel/Dagger.jl/issues/140#issuecomment-735986668