Postgres folds all unquoted column names to lowercase, and doing this
allows that to work nicely without screwing up the queries for the other
database engines.
This removes the following error, caused by the MySQL Dialect using a
'varchar' for all columns over 256 characters in length, instead of
'text' for the largest:
2017/11/17 22:41:33 common_test.go:208: Error 1074: Column length too big for column 'Lua' (max = 21844); use BLOB or TEXT instead
Ensure we don't get silent errors if the Lua code is longer than the
database column, don't leave out the first report from testing, test
fetching multiple reports.
This is done by requiring 'parseTime=true' in the MySQL DSN's (required
by github.com/go-sql-driver/mysql when Scan()ing to time.Time's), and
not forcing update counts to match if rows were updated to what they
already were.