upcall_call_shim_on_rust_stack has redundant call to record stack limit?

[graydon]

Unclear comment in this function in rust_upcall.cpp, but appears to be saying function does some redundant work. Clarify?

[bblum]

This is necessary. When switching to the c stack sp_limit is set to 0, so it must be restored before going back to rust.

The sp_limit needs to be 0 when running on the C stack, so that when crust functions get called, their wrappers don't inadvertently trigger while on the C stack (which can corrupt the "state machine" of saving prev_c_sp and prev_rust_sp, frighteningly).

By the way, I am moving these to rust_task.h as part of the tls push. (#1857)