is None, the next row read is used as the fieldnames.  In the common case,
this means the programmer doesn't need to know the fieldnames ahead of time.
The first row of the file will be used.  In the uncommon case, this means
the programmer can set the reader's fieldnames attribute to None at any time
and have the next row read as the next set of fieldnames, so a csv file can
contain several "sections", each with different fieldnames.

will backport.

figure out what the code is doing

Eliminates the eval() step in the csv module resulting in better
security, more clarity, and a little speed.

The idea is to make successive attempts to coerce the string to
a python type:
    int(s), long(s), float(s), etc.

As a by-product, eliminates a bare 'except' statement.

* Add optional delimiters arg to Sniffer.sniff() which restricts the set of
  candidate field delimiters.

due to using a single module-level namespace where multiple namespaces
were used before.

There *really* need to be tests for the sniffer stuff.  This could
have been avoided.

Skip, please review, and add sniffer tests!