Add LoRa TX timing diagnostics

- Log idle/begin/write/end timing for LoRa TX under SERIAL_DEBUG_MODE
- Document TX timing logs in README

src/lora_transport.cpp

@@ -33,6 +33,18 @@ bool lora_send(const LoraPacket &pkt) {
   if (LORA_SEND_BYPASS) {
     return true;
   }
+  uint32_t t0 = 0;
+  uint32_t t1 = 0;
+  uint32_t t2 = 0;
+  uint32_t t3 = 0;
+  uint32_t t4 = 0;
+  if (SERIAL_DEBUG_MODE) {
+    t0 = millis();
+  }
+  LoRa.idle();
+  if (SERIAL_DEBUG_MODE) {
+    t1 = millis();
+  }
   uint8_t buffer[5 + LORA_MAX_PAYLOAD + 2];
   size_t idx = 0;
   buffer[idx++] = pkt.protocol_version;
@@ -53,8 +65,24 @@ bool lora_send(const LoraPacket &pkt) {
   buffer[idx++] = static_cast<uint8_t>(crc & 0xFF);
 
   LoRa.beginPacket();
+  if (SERIAL_DEBUG_MODE) {
+    t2 = millis();
+  }
   LoRa.write(buffer, idx);
-  int result = LoRa.endPacket();
+  if (SERIAL_DEBUG_MODE) {
+    t3 = millis();
+  }
+  int result = LoRa.endPacket(false);
+  if (SERIAL_DEBUG_MODE) {
+    t4 = millis();
+    Serial.printf("lora_tx: idle=%lums begin=%lums write=%lums end=%lums total=%lums len=%u\n",
+                  static_cast<unsigned long>(t1 - t0),
+                  static_cast<unsigned long>(t2 - t1),
+                  static_cast<unsigned long>(t3 - t2),
+                  static_cast<unsigned long>(t4 - t3),
+                  static_cast<unsigned long>(t4 - t0),
+                  static_cast<unsigned>(idx));
+  }
   return result == 1;
 }