Remove MCDeferred: MCDeferred<MCStack> programState; -> MCStack *programState;

[gc00]

• This used to be: MCDeferred<MCStack> programState;
• MCDeferred is not used anywhere else.
  (A comment there was: "This clever workaround ..." But as others say: "Beware of clever code.")

In the future, we need to make programState a true singleton class:

MCStack programState;

[gc00]

@maxwellpirtle , When I test this code, I'm not seeing a bug. I know that you've seen a bug.
Could you reply here about what target application you are testing, and the flags used with McMini? Thanks.

[maxwellpirtle]

@maxwellpirtle , When I test this code, I'm not seeing a bug. I know that you've seen a bug. Could you reply here about what target application you are testing, and the flags used with McMini? Thanks.

Running

./mcmini ./philosophers_mutex_deadlock 4 0

results in a bug. In the trace printout (when I ran it was trace 23) we see e.g.

...
THREAD STATES
thread 1: exits
thread 1: exits
thread 2: exits
thread 3: exits
thread 4: exits

Thread 1 appears twice

[gc00]

With the latest commits from other PRs, this PR now seems to work.