CFSv2: transient timing output

The second patch here (yes, 2nd, I'll talk about the 1st below) adds the transient timing info already present inside CFSv2's timingInfo struct to the actual timing output file.

@reinhard-prix: I infer that you might have an equivalent commit on a 2010-era branch somewhere, but it apparently never made it to master. ;)

In this patch it always prints the new columns, with defaults of 0 in the non-transient case; let me know if you prefer an if-condition so that the default output format stays the same...? Postprocessing for me would be easier if the columns are always the same, but I don't know if you'd have many scripts that need updating, or if you're always using ComputeFstatBenchmark instead anyway these days. cc @karl-wette on this question, too.

Also I noticed an internal source of confusion in that tauFstat and tauTemplate are averaged over templates, but the transient tauTransMarg and tauTransFstatMap are summed. I've decided to put the patch making the transient timings also averaged before the file-writing patch, so that (from master) we only ever end up with output files from one consistent convention. (I'll just have to make double-sure to remember the different convention in my previous timing results from an older private branch...)