Miscellaneous sched issues

[omor1]

Collection of various issues in different MCA sched components; I'll make a PR to fix these at some point.

• LFQ
  • Currently iterates over local queue twice if local queue is empty
    • Loop in sched_lfq_select should start at i = 1 to avoid this; hierarch_queues[0] is always task_queue
• LHQ
  • Small memory leak by allocating hierarch_queues twice
    • Lines 85–87 should be removed
  • Not sure how this functions when hwloc is disabled or broken, this probably hasn't been tested recently—I didn't see any checks for this, unlike in e.g. LFQ
• LL
  • Minor typo in SDE counter description
    • s/LFQ/LL on line 82
• LTQ
  • ~~Appears that it can lose tasks~~
    • ~Scenario:~
      1. ~~Local task_queue is empty~~
      2. ~~All other task_queue are empty~~
      3. ~~We obtain a heap from the system_queue~~
      4. ~~That heap has 3+ tasks so heap_split_and_steal returns a valid heap in new_heap~~
    • ~~heap gets pushed back to the local task_queue~~
    • ~~new_heap doesn't and all tasks on it disappear~~
    • ~~See lines 217–231~~
    • ~~Either push back new_heap too or just use heap_remove instead of heap_split_and_steal~~
      • ~~heap_split_and_steal is used when stealing a heap from a non-local task_queue so that we don't steal all tasks in that heap~~
      • ~~This isn't a concern when we get a heap from the system_queue~~
      • This is actually more of a problem of documentation and heap_split_and_steal not doing exactly what the name implies
      • The function essentially returns a linked list of two heaps and both get pushed onto the local task_queue.
      • I think that it does cause the heap of tasks with shared inputs to be split into two sub-heaps though.