I see, there is also a INET6_ATON so this may even be feasible. I'm still concerned about performance as every DNS lookup has to interact with the database (checking if it is in gravity, the white- or blacklist).
With SQLite3 this is typically very performant (<1 ms even on a Raspberry Pi Zero). We do a lot to achieve this performance, like preparing all the statements beforehand and keeping the index in process-local (!) memory. This turns out that only a variable binding and step is necessary to get the result. With MySQL being the database (maybe even on a different server), I do not know how the performance would look like.
Having said that, I don't think we have the manpower ourselves to do this soon because, as I said above, the current code is tightly written around SQLite3 and a lot of its features like locally in-memory prepared statements so it would be a lot harder than simply replacing the one function which parses the SQL string. This is not to rule this out, but rather to give a realistic estimate that this seems somewhat out of sight without an external contribution for it.