client: fix Op early destruction

Errors resulting in early Op destruction would erroneously
be escalated to "Not in running evbase worker" exceptions.
Something has already gone wrong, but this would mask the
root cause.

test/testget.cpp

@@ -25,6 +25,16 @@
 namespace {
 using namespace pvxs;
 
+void testEmptyName() {
+    client::Config conf;
+    conf.autoAddrList = false;
+    auto ctx(conf.build());
+    testThrowsMatch<std::logic_error>("Empty channel name",
+                                      [&](){
+                                          ctx.get("").exec();
+                                      });
+}
+
 struct Tester {
     Value initial;
     server::SharedPV mbox;
@@ -592,10 +602,11 @@ void testServerClose()
 
 MAIN(testget)
 {
-    testPlan(67);
+    testPlan(68);
     testSetup();
     logger_config_env();
     const bool canIPv6 = pvxs::impl::evsocket::canIPv6;
+    testEmptyName();
     Tester().testConnector();
     Tester().testWaiter();
     Tester(AF_INET).loopback();