resolve #151

• epics_xy
  • MYA payload is too large; downsample by using l option in myquery
  • Fits are unstable -> use histogram stats instead of a fit
• ftof_tdcadc_p1a, ftof_tdcadc_p1b, and ftof_tdcadc_p2
  • problem: memory-resident histograms are too high resolution
    • out of memory occurs after 419 files * 6 sectors * 3 timelines = 7532 histograms, each with ~16,000 bins
    • increase memory allocation pool from 1GB to 2GB makes RG-C work
    • default number of threads is 8, so this is a 16 GB allocation... maybe too much
    • can't really reduce the histogram range or binning, since the bin width must remain as is and the range is set such that outliers are captured (e.g., early RG-A sp18)
  • solution:
    • new function HistoUtil.zoomHisto truncates the "empty" parts of a histogram, focusing on the region with filled bins only; for a test histogram, this zoomed histogram has around 45% of the bins
    • if we still use too much memory for large datasets, run those timelines with the new option set $TIMELINE_JAVA_OPTS_HIGHMEM, which will increase the mempool allocation (defined in this PR, but not yet used)
• FTOF ftof_tdcadc_p2: now that they work, we see the fits are bad
  • stabilized by fixing initial RMS (as p1a and p1b do) and reducing the fit range