feature: binned histograms

Histograms could be done in numerous ways. Here are some thoughts:

• like most of xsv, should operate over huge tables with a single pass
• .idx files could store metadata helpful for binning? eg, min and max values per column
• "chose N evenly spaced numerical bins" seems to require more than one pass (or keeping all values in memory). Keeping a tree of round-sized bins and merging them when the tree gets too big would avoid that
• logarithmic bins
• power-of-two or power-of-1024 (eg, for file sizes)
• binning of strings or decimal numbers by prefix
• "other" / "NaN" / null bins
• csv/tsv output by default, then a separate mode like xsv table to pretty print bars