Collects the products generated during the compilation of a project.

It has two previous results. The first is to be used by the projects that depend on this compilation and are being compiled during this run. This result is used because it contains class file locations referring to both classes directories and because these two classes directories will be in the dependency classpath of dependent compilations.

The latter previous result is used for future client compilations. Future compilations will not see the readOnlyClassesDir and newClassesDir in this abstraction, but instead see an aggregated read-only classes directory. newClassesDir will become this read-only classes directory by the time of that compilation because all compilation products (class files, semanticdb files, sjsir files, tasty files, nir files, etc) in the current readOnlyClassesDir that are not in newClassesDir will be copied over. Therefore,the result that is used for compilation will have all class file references rebased to newClassesDir.

We use different previous results because we want to remove the cost of copying class files from readOnlyClassesDir to newClassesDir before starting dependent compilations. This scheme allows us to be fast and perform as much IO work in the background as possible.