#708: Pass subphase timings through to ProcStats, and write them out

[PhilMiller]

Fixes: #708

[PhilMiller]

I think we should close #708 with this, since further work on models and/or strategies that make effective use of the subphase timing data is a pretty open-ended project.

[PhilMiller]

@ppebay Have a look at the format in which I'm writing out the subphase times. Will that pose a problem for the analysis tools?

[lifflander]

@PhilMiller What happened here? Why is MPI intercept code on this branch?

Can you write a Doxygen comment documenting the new format (and current one also)?

[PhilMiller]

Rebased to clear up the integration issues.

Wrote up a full description of the stats file output in the comments.

[lifflander]

Looks good to me.

[lifflander]

This seems to be failing Travis CI sometimes

[codecov]

Codecov Report

Merging #826 into develop will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##           develop     #826      +/-   ##
===========================================
+ Coverage    80.49%   80.50%   +0.01%     
===========================================
  Files          353      352       -1     
  Lines        11207    11188      -19     
===========================================
- Hits          9021     9007      -14     
+ Misses        2186     2181       -5     

Impacted Files	Coverage Δ
src/vt/vrt/collection/types/migratable.h	50.00% <ø> (ø)
src/vt/vrt/collection/balance/elm_stats.impl.h	92.68% <100.00%> (+0.18%)	⬆️
tests/unit/atomic/test_atomic.cc	100.00% <0.00%> (+1.31%)	⬆️