Completed refactor

- Writing to CSV is now done with openCSV calls, and client is
  responsible for lifecycle of injected dependencies (closes #1)
- Because of the openCSV calls, commas within a field are now allowed
  on output, as is enclosing a field in quotes (closes #3)
- Because reading had already been rewritten using openCSV calls,
  reading already could handle quoted fields that contained commas; we
  simply didn't have a test for it until now (closes #2)