sql.py: put_table(): Take a dict mapping out to in cols instead of separate in and out cols lists

sql.py: mk_select(): Joins: Reversed order of left_col and right_col in the joins dict as well, so the joined table's columns are the keys

sql.py: mk_select(): Joins: Reversed order of left_col and right_col so the column of the table being joined is first, to match the form of a WHERE clause

sql.py: mk_select(): Support joins

sql.py: mk_select(): Accept a list of tables to join together (initial implementation just uses the first table)

sql.py: mk_select(): Support ORDER BY clause. By default, order by the pkey, since PostgreSQL apparently doesn't do this automatically (and this was causing some staging table tests to fail).

bin/map: In debug mode, print the row # and input row just like in error messages

bin/map: verbose_errors also defaults to on in debug mode

sql.py: add_row_num(): Make the row number column the primary key

csv2db: Use new sql.cleanup_table() to map NULL-equivalents to NULL. Consider the empty string to be NULL.