Early work to make it possible to call the interpreter from ReadyToRun code

[cshung]

This prototype implements the most basic version of ready to run support to let the interpreter to work with ready to run code. It is far from complete. Right now, it can

• Call any ReadyToRun code from the interpreter, by choosing not to use InterpreterPrecode in method.cpp, and
• call static, parameterless method from ReadyToRun code to interpreter, using a few redirections.

This is for Windows, x64 only.

1. When ReadyToRun code calls a method for the first time, it will end up in ExternalMethodFixupWorker to determine eventually where that will go.
2. If the runtime decide it should be interpreted, it need to do some preparation work:

2.1 Create the InterpreterMethodInfo object and make it available through the newly introduced InterpreterImport redirection cell, and
2.2 Redirect the call to a InterpreterStub that will do two things:

• Move the address of the InterpreterMethodInfo object from the InterpreterImport redirection cell to rcx, and then
• Jump to the interpreter entry point, which is made available though an eager import that we prepared earlier during assembly load for the ready to run image.

With that the interpreter will get the interpreter method info and work on it and return to the original caller.

The AppContext changes are just so I can test making these calls - the math is there just to avoid optimization.

To test this out, simply run any application (since it will run AppContext setup and fail there regardless) with the same set of environment variables as before but turn on DOTNET_ReadyToRun.