* handle CR as a newline (and perhaps also others from 
  http://en.wikipedia.org/wiki/Newline)
* a separate CSV format, that handles quoted strings properly,
  based on boost.tokenizer:
  http://stackoverflow.com/questions/1120140/csv-parser-in-c
  non-numeric fields should be read as NaNs. Options: first-line-header.
* is Rigaku RIG just another common extension for rigaku_dat? If yes, add it.
* implement GSAS format (see gsas.cpp for comments) 
* PANalytical XRDML
* consider adding JCAMP format (Standard NMR Data Format)
* text format: handle other numeric formats (e.g. ',' as a decimal separator)
* (?) support URIs (using libcurl, libfetch or boost:asio)

