Fix tuple syntax when using key sequence in distinct()

[Kirkman]

I have noticed in the past that Table.distinct() does not deduplicate when the key is a list or sequence of column names.

Code like this (wrongly) results in no rows being dropped ...

table = table.distinct(['PARCELID', 'SITEADDR'])

... whereas code like this does successfully drop duplicate rows:

table = table.distinct(lambda row: str(row['PARCELID']) + str(row['SITEADDR']))

I believe the cause is line 31 of agate/table/distinct.py, which constructs the row's key for de-duplicating:

k = (row[j] for j in key)

I think the parentheses here create are creating a generator expression, not a tuple. k is always unique, and no de-duplication ever occurs.

My suggested fix is to wrap it with tuple(...):

k = tuple(row[j] for j in key)