Thread cf & must edges

[ThomasHaas]

This PR adds proper control-flow dependencies between spawned threads and their spawners/creators.

• Each thread's entry event is now a ThreadStart event that may have a ThreadCreate event as a creator. If creator == null, the thread is executed unconditionally (e.g, the main thread, init threads, and all threads in litmus tests).
• The control-flow encoding now allows the execution of ThreadStart only if its associated creator (if any) gets executed.
• BranchEquivalence has been reworked to respect this new control-flow and consider the control-flow relationship between a thread and its creator.
• Synchronization events between spawning thread and spawned thread have been simplified.
• We now generate must-rf edges for synchronizing events.

Minor side changes:

• SCCP now properly recognizes AbortIf (true) for dead code detection
• SCCP now eliminates trivial casts that do not change types (e.g., i64 to i64 casts).
• Thread.getEntry and Thread.getExit now return ThreadStart resp. Label.

[hernanponcedeleon]

what happens if I use the CLI option to force thread creation? If that is the case do we call setMayFailSpuriously()?

[ThomasHaas]

Yes, ThreadCreation will spawn threads and set the option based on whether thread creation is forced to succeed.
At some point, I will probably get rid of this flag again and instead let ThreadCreation generate a non-deterministic choice between success and failure like so:

// compiling "int success = pthread_create(&t, NULL; worker, arg);" to
if (*) {
    ThreadCreate(worker, arg);
    store (&t, TID);
    success = 1;
} else {
    // store (&t, -1);
    success = 0;
}

Then ThreadCreate/ThreadStart can always be assumed to succeed if executed.

[hernanponcedeleon]

LGTM