gh-138122: Implement frame caching in RemoteUnwinder to reduce memory reads (#142137)

This PR implements frame caching in the RemoteUnwinder class to significantly reduce memory reads when profiling remote processes with deep call stacks.

When cache_frames=True, the unwinder stores the frame chain from each sample and reuses unchanged portions in subsequent samples. Since most profiling samples capture similar call stacks (especially the parent frames), this optimization avoids repeatedly reading the same frame data from the target process.

The implementation adds a last_profiled_frame field to the thread state that tracks where the previous sample stopped. On the next sample, if the current frame chain reaches this marker, the cached frames from that point onward are reused instead of being re-read from remote memory.

The sampling profiler now enables frame caching by default.

Python/ceval.c

@@ -2288,6 +2288,16 @@ clear_gen_frame(PyThreadState *tstate, _PyInterpreterFrame * frame)
 void
 _PyEval_FrameClearAndPop(PyThreadState *tstate, _PyInterpreterFrame * frame)
 {
+    // Update last_profiled_frame for remote profiler frame caching.
+    // By this point, tstate->current_frame is already set to the parent frame.
+    // Only update if we're popping the exact frame that was last profiled.
+    // This avoids corrupting the cache when transient frames (called and returned
+    // between profiler samples) update last_profiled_frame to addresses the
+    // profiler never saw.
+    if (tstate->last_profiled_frame != NULL && tstate->last_profiled_frame == frame) {
+        tstate->last_profiled_frame = tstate->current_frame;
+    }
+
     if (frame->owner == FRAME_OWNED_BY_THREAD) {
         clear_thread_frame(tstate, frame);
     }