Use secondary storage during OST generation

[Scott Wells]

Right now OST generation is very memory-intensive because it holds the results of all API calls in-memory until the OST is fully-generated. In large(-ish) orgs, this can lead to a need for an equally large (sometimes huge) max heap for the IDE process itself. Instead API results should be written into secondary storage efficiently keyed for later retrieval as the information is being threaded together into the OST class files themselves.

[Scott Wells]

I don't think this is really necessary anymore. At some point I did some work to keep things relatively lean during OST generation against all but the absolute largest orgs (which admittedly some people work against!) and haven't really heard about chronic issues with memory during OST generation in a long time.