Tasks: Fix Compiler Errors

Tasks: rewrite task scheduler and pools to use TBB

TODO: test performance
* Local queues in depsgraph
* Grain size in parallel range
* Iterators chunk size calc

Tasks: don't pass thread_id always, instead go through accessor

We should move away from using thread IDs and instead store any thread local
data in the TaskParallelTLS. Getting the thread ID will also be more expensive
with TBB and is not so well defined.

Tasks: remove scheduler struct, there is only a global one now

Tasks: use TBB for parallel range

Tasks: split task.c into task_pool.cc and task_iterator.c

Task: change func_finalize to func_reduce and func_free

In preparation of using TBB.

Task: remove local queue and scheduling mode optimizations

Hopefully no longer needed with TBB, but needs to be tested.

Task: don't use suspended task pool for iterators

There is no apparent reason for this, and it has extra overhead.

Tasks: move priority from task to task pool