Fix end-of-stack native resume when skip_native_resume is set

When skip_native_resume is true but jit_detected is false (PC not in JIT
region at sample time), the end-of-stack check fell through to
PROG_UNWIND_NATIVE. The native unwinder could then encounter the JIT
anonymous mapping and re-enter the Ruby unwinder, pushing a JIT frame
in the wrong stack position.

Check skip_native_resume alongside jit_detected at end-of-stack to
prevent resuming native unwinding when the flag is set.

Author: dalehamel

support/ebpf/ruby_tracer.ebpf.c

@@ -519,7 +519,8 @@ static EBPF_INLINE ErrorCode walk_ruby_stack(
       // We have processed all frames in the Ruby VM and can stop here.
       // If JIT was detected without FP, the PC is still in the JIT region
       // and native unwinding would fail, so we stop.
-      *next_unwinder = record->rubyUnwindState.jit_detected ? PROG_UNWIND_STOP : PROG_UNWIND_NATIVE;
+      *next_unwinder = (record->rubyUnwindState.jit_detected || rubyinfo->skip_native_resume)
+                           ? PROG_UNWIND_STOP : PROG_UNWIND_NATIVE;
       goto save_state;
     } else {
       // If we aren't at the end, advance the stack pointer to continue from the next frame