switch back to SSD

[ctm]

Provision and use a new ec2 instance that uses SSD, rather than magnetic.

I went with magnetic, figuring everything would be cached so it didn't really matter. For the server itself, that may indeed be the case. I certainly haven't seen any sign of slowdown. However, deployments take longer and ssh'ing in takes longer. Most likely that's just a matter of little used utilities' memory being forgotten.

Eventually, when we do blue green deploys with switchover being done by load balancers, this won't be an issue, but that's not happening in the near future. In the meantime, it's not important enough to merit a high priority label, but I don't want to get used to the new slow deployments and think that there's nothing that can trivially speed them up, because there is.