fix bug in job step gres allocation logic

  for logging purposes. no real change in logic

  state, which causes serious problems now that they are different.
Add a node_in_use bitmap to step's gres state.
Improve gres step de/allocation logic

if the cluster you are looking at isn't responding and you look at another cluster the grid is now set up correctly

fixed individual to only really look at the 'i' instead of all on individual.  And also made it so acct was accepted more places

    process threads on a last-in first-out basis and the high thread count was
    causing unexpectedly high delays for some RPCs.

let job/step specify gres requirement with suffix of M (mega)

and add support for topology going away on an existing gres

 (the plugin is optional).
Let resource counts be specified with a "K" or "M" suffix.

BLUEGENE - Added option to scontrol & sview to resume blocks in an error state instead of just freeing them.