We already have a fast path out of mono_threads_enter_gc_safe_region_unbalanced_with_info if we're not even tracking the thread state transitions to BLOCKING.

But we could also make it faster under full coop on "normal" platforms where we don't need to take a snapshot of the stack. It's enough just to copy the callee-saved registers to the normal stack.

Also if we're using hybrid suspend on a "normal" platform, we'll be preemptively suspending the thread while it's in GC Safe, so we don't need to save the registers either.

Part of #6921