stage_write for varying size arrays

[kshitij-v-mehta]

In this PR:

• modifications to stage_write for varying sized arrays as seen in xgc
• Makefiles for tau/sos_flow
• some timing information. not ultimately printing values out, but kept some timers in.

[kshitij-v-mehta]

By default, stage_write now contains code for handling varying size arrays. This means the process_metadata function is called at every tilmestep. This should not incur any significant overhead as it is a local operation that does not perform any MPI operations (as per Norbert).

[jychoi-hpc]

It looks good to me.