Optimize buffer operations for better upload performance

- Replace byte-by-byte copies with memcpy (10-100x faster)
- Increase SD write chunk size from 4KB to 16KB
- Use static buffer for SD writes to reduce stack usage
- Remove unnecessary yield() from handleClient loop

The byte-by-byte circular buffer operations were a major bottleneck.
Using memcpy with proper wrap-around handling significantly improves
throughput.

src/activities/network/CrossPointWebServerActivity.cpp

@@ -327,7 +327,6 @@ void CrossPointWebServerActivity::loop() {
       // in chunks and each handleClient() call processes incoming data
       for (int i = 0; i < HANDLE_CLIENT_ITERATIONS && webServer->isRunning(); i++) {
         webServer->handleClient();
-        yield();  // Allow other tasks to run between iterations
       }
       lastHandleClientTime = millis();
     }