Detect and remove UTF-8 BOM

Created by: doekman

The bug is illustrated with files from this zip

example.csv is a CSV file, saved with UTF-8 with BOM marker and LF linefeeds (the webapp highrise uses this format).

When running ./setup.sh the file is imported into the user-database, and everything seems fine:

csvsql 1.0.2
Python 2.7.15
ProductName:	Mac OS X
ProductVersion:	10.13.6
BuildVersion:	17G3025
NOTICE:  table "example" does not exist, skipping
DROP TABLE

However, when running ./test.sh, the second command fails (and the first and third succeed):

----[ select * ]--------------------------------------------
   FullName   |        Email        
--------------+---------------------
 Test User    | test@example.com
 Another User | another@example.com
(2 rows)

----[ select FullName, Email ]------------------------------
ERROR:  column "fullname" does not exist
LINE 1: SELECT FullName, Email from example;
               ^
HINT:  Perhaps you meant to reference the column "example.FullName".
----[ select [BOM-marker]FullName, Email ]------------------
   FullName   |        Email        
--------------+---------------------
 Test User    | test@example.com
 Another User | another@example.com
(2 rows)

This is because the first column name is prefixed with the BOM marker, and is demonstrated by the third command. When you copy/paste the output into an editor that shows control characters (TextMate), you see the name of the first column is actually `U&"\FEFFFullName"

When the header columns in the CSV file are double-quoted ("FullName","Email"), the behaviour is different but also undesirable, but I haven't checked that.

The error message could be nicer (non-printable characters could be shown escaped as I did in the third example), but I'm not sure which software part that would have to do...

From csvkit, I expect the UTF-BOM marker to be detected and removed.