type-context: revert client disconnect workaround

This reverts commit 196e6fc from
bitcoin-core#118 which added a
workaround which is no longer needed after
315ff53 from
bitcoin-core#160

When the workaround was introduced it prevented segfaults that happened when an
IPC client disconnected during a long-running asynchronous call. But it
prevented them by leaking server objects, and these leaks now sometimes cause
the new "disconnecting and blocking" unit test introduced in bitcoin-core#160 to hang.
Since the workaround is no longer necessary, revert it now to fix the test
hangs.

The problem with test hangs was reported:

- Sjors/bitcoin#90 (comment)
- https://github.com/Sjors/bitcoin/actions/runs/15966265407/job/45027248310?pr=90
- Sjors/bitcoin#90 (comment)
- https://cirrus-ci.com/task/4999408900636672

And the fix was posted:

- Sjors/bitcoin#90 (comment)

Author: ryanofsky

include/mp/type-context.h

@@ -69,7 +69,6 @@ auto PassField(Priority<1>, TypeList<>, ServerContext& server_context, const Fn&
                 const auto& params = call_context.getParams();
                 Context::Reader context_arg = Accessor::get(params);
                 ServerContext server_context{server, call_context, req};
-                bool disconnected{false};
                 {
                     // Before invoking the function, store a reference to the
                     // callbackThread provided by the client in the
@@ -101,7 +100,7 @@ auto PassField(Priority<1>, TypeList<>, ServerContext& server_context, const Fn&
                     // recursive call (IPC call calling back to the caller which
                     // makes another IPC call), so avoid modifying the map.
                     const bool erase_thread{inserted};
-                    KJ_DEFER({
+                    KJ_DEFER(if (erase_thread) {
                         std::unique_lock<std::mutex> lock(thread_context.waiter->m_mutex);
                         // Call erase here with a Connection* argument instead
                         // of an iterator argument, because the `request_thread`
@@ -112,24 +111,10 @@ auto PassField(Priority<1>, TypeList<>, ServerContext& server_context, const Fn&
                         // erases the thread from the map, and also because the
                         // ProxyServer<Thread> destructor calls
                         // request_threads.clear().
-                        if (erase_thread) {
-                            disconnected = !request_threads.erase(server.m_context.connection);
-                        } else {
-                            disconnected = !request_threads.count(server.m_context.connection);
-                        }
+                        request_threads.erase(server.m_context.connection);
                     });
                     fn.invoke(server_context, args...);
                 }
-                if (disconnected) {
-                    // If disconnected is true, the Connection object was
-                    // destroyed during the method call. Deal with this by
-                    // returning without ever fulfilling the promise, which will
-                    // cause the ProxyServer object to leak. This is not ideal,
-                    // but fixing the leak will require nontrivial code changes
-                    // because there is a lot of code assuming ProxyServer
-                    // objects are destroyed before Connection objects.
-                    return;
-                }
                 KJ_IF_MAYBE(exception, kj::runCatchingExceptions([&]() {
                     server.m_context.loop->sync([&] {
                         auto fulfiller_dispose = kj::mv(fulfiller);