Make sure we test issue deduplication

The latest database and I/O schema versions have added support for "issues" and "incidents". Out of those two "issues" required a potentially-disruptive change, because they have two columns in their primary keys: id and version. As such, the deduplication mechanism should handle them differently from other objects, i.e. allow keeping issues with the same id field value, but different version values in the database at the same time. This is different from the rest of the objects, which all can differ in their id field only.

Make sure we have tests in kcidb/test_db.py that:

• load() two issues with the same id/version pair and check that only one ends up in the database (using either the query() or the dump() method);
• load() two issues with the same id, but different version field values, and check that both are there in the database (using the same method);
• load() two issues with the same version, but different id field values, and check the same;
• load() two issues with both id and version different, and check the same.

Either prove tests like that already exist, or make sure they don't and add your own.