Testbench throughput and profiling updates

Tweak testbench and arch configuration such that it's easy to centrally produce a flamegraph of only the event loop.

Goes with lhcb-core/LHCbNightlyConf!1212 (merged) and lhcb-datapkg/PRConfig!419 (merged)

MooreScripts/python/MooreScripts/testbench/emulator.py

@@ -228,16 +228,16 @@ class Task:
             raise RuntimeError("use_perf() must be called before load()")
         if shutil.which("perf") is not None:
             self._perf_data = Path(f"{self.utgid}.perf.data").absolute()
-            if control:
-                self._perf_control = Path("perf_control.fifo").absolute()
-                try:
-                    os.mkfifo(self._perf_control)
-                except FileExistsError:
-                    pass
         else:
             log.warning(
                 "Requested to run perf but no executable found. Continuing without profiling."
             )
+        if control:
+            self._perf_control = Path("perf_control.fifo").absolute()
+            try:
+                os.mkfifo(self._perf_control)
+            except FileExistsError:
+                pass
 
     async def __aenter__(self):
         pass