Or (shameless plug) use tokudb. We have many online schema change features that work within mysql. You can add and remove columns while the database is running, without resorting to complicated tricks with triggers. You just add the column, it completes immediately, and you start using it right away. The work of changing existing rows happens in the background, in a transactionally consistent manner, so you never see the old schema again. It's pretty sweet.
FWIW, I often look at tokudb with envy, but there is no way I will switch to MySQL from PostgreSQL for it: there are so many other things I'd lose that I would rather put up with my other issues. That said, and I totally admit it was now over a year ago that I was looking into various database implementation companies (and thereby might have you confused with a differen company), it seems to me like your technology would be amazing if used as a standalone index (as opposed to also storing the heap), and adding indexes to PostgreSQL is both simple and "part of the point of using it in the first place".
(I also would have expected the PostgreSQL market to be better for this kind of thing, although I'm not in this business, so I'd love to understand where I'm wrong: the general argument being that relying on a third-party commercial index is a much lower commitment--and thereby a simpler sell--than relying on a third-party commercial storage engine, in addition to how most companies using MySQL seem to either know so little about database technology that their criteria was simply "popularity" or, alternatively, so much about database technology that to them MySQL is perfectly fine and they don't really need your solution.)
For the moment, I can say that MySQL has a nicely defined storage engine API which is why we release that, and that if you need something lower-level, please come talk to us because that's not the only thing we develop.
However, part of the point of storing things in a write-optimized data structure is that you can afford to store "the heap" directly in the index, and get good performance on range queries. Anyway, if you're interested, email us and we can talk about how we might be able to help.
TokuDB is interesting but priced a little outside what I'd be willing to pay. For the same price I can throw more machines at the problem and use Cassandra.
