It would make the rocket too heavy. To add another engine, they would have to make the rocket larger in diameter, increasing the weight by a fair amount (they currently have the 9th engine in the middle, so it would have to get moved to the ring with the rest). Reducing the power of the center engine would also reduce the amount of payload to orbit.
Rockets that can reach orbit have very bad weight to payload ratios (Falcon 9 is ~2.5% takeoff weight to LEO) so there isn't much room for dead weight.
It's actually more about complexity. They could simply replace the center engine with a lower power design and still get plenty of mass to orbit. The issue then is that you have to design a completely new engine and have 2 different engines in your first stage. More complexity generally equals more risk, and historically SpaceX has sought to maximize simplicity for the sake of reliability and rapid iteration.
And besides, it would be a waste. Landing with the engines they have is "just" a software problem. (Technically a control system problem.) Assuming they can make it work consistently, the save the added cost and inefficiency of a separate landing engine for all future launches.
It has eight other engines to fall back on, and engine restarts are already normal part of the return flight. Centre engine failure before the final burn starts would certainly be survivable. It's just a question of the software using two or three engines in a symmetrical pattern for a 1/2 or 1/3 of the length of the 1-engine burn. I'm not sure if they could also fire one engine asymmetrically and use momentum and the secondary positioning system of cold thrusters to counter the asymmetry.
I don't think that the software is enabled yet -- SpaceX usually press forward with the main objective first, and flesh out the contingencies later (for example, abort hardware was present during the last failed launch, and it had been tested before, but the software wasn't enabled, and they lost the vehicle).
Landing is already severely constrained by the minimum thrust of the engine though. At a minimum using two would make the tolerances much, much tighter. Remember it wouldn't be double the thrust but rather several times the net thrust, since the downward thrust from gravity would stay the same.
Because the return part of the mission relies on nine identical engines right now. So if one fails it is not a problem. But if you replace the central one with a different one that becomes critical for the return mission then you have made matters worse.
> Because the return part of the mission relies on nine identical engines right now. So if one fails it is not a problem.
Is that true? Doesn't the landing have to use the center engine - any other single engine would be asymmetric, and multiple engines would have too high a thrust.
Yes, if the center engine fails, the landing fails. There's no way to recover from that.
But the great thing about SpaceX's system is that it starts with a cheap expendable rocket, then makes it recoverable. You lose one? Who cares, build another one. They could have a 50% success rate recovering the first stages and it would still make things amazingly cheaper.
For future reference, the return part is made up of 3 burns. The first two use three engines, and the landing burn uses just the one. Coming back, it's obviously much lighter without all the propellant and 2nd stage/payload weight, so it needs much less thrust to reverse course. See http://flightclub.io for more info.
Rockets that can reach orbit have very bad weight to payload ratios (Falcon 9 is ~2.5% takeoff weight to LEO) so there isn't much room for dead weight.