Draft: Introduce smaller RTS entrypoint
As described in Note [Simple main], this is an attempt to pick off some low-hanging fruit in the time necessary to link executables. Specifically we try to minimize the amount of work necessary to compile C stub produced when linking an executable. On my machine the time necessary to compile this stub from from more than 60ms to less than 20ms.
This cost was originally noted in #16822.
Edited by Ben Gamari