Reading CSVs with UTF8 BOM

[mcoady]

This might be more of a cassava issue (there's a decade-old issue there), but when I've used cassava in the past I've had to do this "fix" when reading certain csv files (usually when making them in Excel). But might be useful to add the change into dataframe from a UX perspective maybe? Basic fix here.

Current behaviour (due to cassava)

dataframe> D.readCsv "cdr.csv"
--------------------
*** Exception: <stdout>: hPutChar: invalid argument (cannot encode character '\65279')

Behaviour after change

dataframe> D.readCsv "cdr.csv"
-------------------
period | percentage
-------|-----------
 Int   |   Double
-------|-----------
-1     | 0.38
0      | 0.76

Lazy

I see the lazy version does load and picks these up as valid characters, but not sure if that should be the case. Had a look at trying to do a similar "fix" but would need to research the mechanics more.

dataframe> CSVL.readCsv "cdr.csv"
----------------------
´╗┐period | percentage
----------|-----------
   Int    |   Double
----------|-----------
-1        | 0.38
0         | 0.76

[mchav]

The lazy API probably needs to be rewritten for parity with the eager API. I think a lot of that work is blocked by the Parquet work that'll happen this summer. The UTF-BOM fix looks good to me. You can open a PR against this issue.