coredump when running with tensorflow 2.15.0

**Describe the bug (描述bug)**

## Compiled successfully, But coredump when initialize a new tensorflow tensor instance.
feels like sth wrong with protobuf?

### STACK TRACE:

2023-12-28 13:04:36.782611: I external/local_tsl/tsl/cuda/cudart_stub.cc:31] Could not find cuda drivers on your machine, GPU will not be used.
2023-12-28 13:04:38.906209: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:9261] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2023-12-28 13:04:38.906258: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:607] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2023-12-28 13:04:39.205097: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1515] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
123server startbefore start
F20231228 13:04:40.017664 3244 global.cpp:309] external/com_google_protobuf/src/google/protobuf/generated_message_reflection.cc:3012 CHECK failed: file != nullptr:
*** Check failure stack trace: ***
@ 0x7fc6fa71fd33 google::LogMessage::Fail()
@ 0x7fc6fa72724c google::LogMessage::SendToLog()
@ 0x7fc6fa71f8a0 google::LogMessage::Flush()
@ 0x7fc6fa72c78f google::LogMessageFatal::~LogMessageFatal()
@ 0x7fc6f989a5a3 brpc::BaiduStreamingLogHandler()
@ 0x7fc6fc5b9401 google::protobuf::internal::LogMessage::Finish()
@ 0x7fc6fc4f817b google::protobuf::(anonymous namespace)::AssignDescriptorsImpl()
@ 0x7fc6f77b5817 (unknown)
@ 0x7fc6fc4f7ea7 google::protobuf::internal::AssignDescriptors()
@ 0x7fc6fc4b23dc google::protobuf::(anonymous namespace)::RetrieveOptions()
@ 0x7fc6fc486c36 google::protobuf::(anonymous namespace)::FormatLineOptions()
@ 0x7fc6fc489c93 google::protobuf::ServiceDescriptor::DebugString()
@ 0x7fc6fc48c929 google::protobuf::ServiceDescriptor::DebugString()
@ 0x7fc6f99648cb brpc::ProtobufsService::Init()
@ 0x7fc6f99651ee brpc::ProtobufsService::ProtobufsService()
@ 0x7fc6f98e392b brpc::Server::AddBuiltinServices()
@ 0x7fc6f98ea913 brpc::Server::StartInternal()
@ 0x7fc6f98ec2d6 brpc::Server::Start()
@ 0x7fc6f98ec4b7 brpc::Server::Start()
@ 0x557ba946a2e0 main
@ 0x7fc6f7744a90 (unknown)
@ 0x7fc6f7744b49 __libc_start_main
@ 0x557ba9469c75 _start
Aborted (core dumped)

**To Reproduce (复现方法)**

**Expected behavior (期望行为)**

**Versions (各种版本)**
OS: linux x86, ubuntu 23.04 in docker
Compiler: bazel g++ (Ubuntu 12.3.0-1ubuntu1~23.04) 12.3.0
brpc: master
protobuf: 3.21.12-1ubuntu7
tensorflow : 2.15.0.post1
c++ link dynamic lib

**Additional context/screenshots (更多上下文/截图)**

Start by reproducing the crash in the stated Ubuntu 23.04 Docker setup with brpc master, TensorFlow 2.15.0.post1, protobuf 3.21.12, and dynamic linking. Trace initialization from brpc::Server::Start through AddBuiltinServices and ProtobufsService::Init, using the generated_message_reflection.cc:3012 failure as the checkpoint; done means startup succeeds without the coredump.

Written by the indexing model from the issue text.