* ctau changes:
  k rm 4os
  k rm batching at all levels
  k option (default off) to print pre-therm value for each RV.
  k rm unused stat funcs
  k vectorize ... stats routines will be simpler now.  save them somewhere!!
    :(..8
  k bounds-checking routines à la tauint H.
  k how does var tauint behave with increasing N (head -N of same cump data)?
  k rm tol from all C/py versions of tauintflat.
  k ihisto data structure and methods
  k re-order mcmc_rvs.c code to match .h.
  k subscript *all* mu's and sigma's.
  k error-bar plots for YNbars?
  k make e.o.e.b. plots for eta=0.0, 0.9
  k check empirical bias of chat(t)'s for various t and eta: it's low.
  k pyautocorreta
    * cump $eta 40 1000 1000 > c
    * pyautocorreta c $eta 100  > a
    * pgr -n -l -formats 's . . . .' {etc.} a
  k how does *bias* of u^2 behave with respect to batching?
  k emphasize that u is a rough estimator.  "define" that. :)
  k Consider simplifying / eliminating / xreffing fwd the big prob-stt table.
  k within 10% is fine -- it's an error bar.
  k hattauint looks like an underestimator of tauint; god damn.
    * The 1+t/N term doesn't help.
	* What is the expected value of hattauint??
	* Is chat(t) unbiased for c(t)?  What is the expected value of the former?
  n fix up formatting in eta_corrected_meanvar_batch.py:  this will get
    used again. :)
  k include true values in the B table.
  k fW
  k mcrcm code: tauint-correct for all rvs ... or optional.
    or only tauint these:  H, r2, lmax/N, fI (except not?), fS, fW, xi