When using the column hashing feature for a Postgres connector (but assume this applies to other connectors also), if the value is not NULL we'd expect a hash / string, and if the value is NULL then we'd expect NULL.

However, NULL values are indeed being hashed (and we now can't rely on queries such as `... WHERE "<column-name>" IS NULL`).

All NULL values are hashed to the same string, but then we have to transpose `WHERE "<column-name>" IS NULL` to `WHERE "<column-name>" = '<hashed-null-value>'` in queries (meaning that each query environment needs to know the value of `<hashed-null-value>` which frankly is a huge pain).