Fix TSAN data race in ManagedObjectStore::clearAllObjects

In ManagedObjectStore::clearAllObjects(), boost::asio::post was invoked
with a single argument (the lambda callback) instead of passing
io_context_main_thread_. In Boost.Asio, 1-argument post() posts to
boost::asio::system_executor (the background thread pool). This caused
the priority queue reset to execute asynchronously on a background thread
concurrently with io_context thread access (e.g., GetPriorityQueueSize),
causing a ThreadSanitizer data race.
Google-Bug-Id:511115471
PiperOrigin-RevId: 966582297
Change-Id: Iaf820ef2e516016f541c6443ebb397810f75b25b