Avoid double inserting TableCache in multiple threads

If we use the DB::Get interface to read the same data in the SST in multithreading.in table_cache.cc, we use the Lookup interface to check whether the handle exists, the situation may occur: if we have two threads A,B. Thread A is scheduled before thread B. When thread A calls Lookup but does not insert handle into the cache, thread B calls Lookup, which results in * handle == nullptr. Should we add a mutex to the Find interface to prevent a double cache, or better yet, fragment the mutex(use file_number)