validates uniqueness of should take advantage of the database

Currently, case-insensitive validations use ruby to lower case a string on one side, and the database to lower case on the other. That results in a question that looks like this:

"SELECT     \"topics\".\"id\" FROM       \"topics\" WHERE     (LOWER(\"topics\".\"title\") = E'я тоже УНИКАЛЬНЫЙ!') LIMIT 1"

This is fine, except that sometimes Ruby does not know how to lower case the same way the database does. In fact, with 1.9.2, the "test_validate_case_insensitive_uniqueness" test fails because ruby 1.9 does not downcase the string the same way ruby 1.8 does.

This patch switches the validation to use LOWER() on both sides of the comparison, and avoid what Ruby thinks.