Large schemas with hundreds of top level elements can take minutes to compile.

Even if we speed this up massively, it will still be expensive.

When we package a schema into a jar, we should package the compiled processor for it as well, so that this can be reloaded instead of compiling from the DFDL schema source.