diag: add startup-timing trace and /api/diag/timing endpoint

bbctrl.Trace records monotonic-anchored events from process start.
Ctrl, Comm, the Web layer and __init__ are instrumented so a single
GET /api/diag/timing returns a full timeline of import, controller
init, AVR connection, first websocket, and first GET /. The
restart-timing.js client posts performance.now() marks back so the
browser side can be aligned in the same view.

Used to drive the cold-boot optimisations that reduce listen latency
on the Pi by ~8s.

src/py/bbctrl/__init__.py

@@ -36,6 +36,13 @@ import datetime
 
 from pkg_resources import Requirement, resource_filename
 
+# Trace must be imported before the rest of bbctrl so its monotonic
+# anchor is the earliest reasonable point and so import-time costs of
+# heavy submodules (camotics gplan.so, sockjs, tornado, etc.) are
+# attributable in /api/diag/timing.
+import bbctrl.Trace as Trace
+Trace.mark('imports.bbctrl.start')
+
 from bbctrl.RequestHandler import RequestHandler
 from bbctrl.APIHandler import APIHandler
 from bbctrl.FileHandler import FileHandler
@@ -64,6 +71,8 @@ import bbctrl.v4l2 as v4l2
 import bbctrl.Log as log
 import bbctrl.ObjGraph as ObjGraph
 
+Trace.mark('imports.bbctrl.end')
+
 
 ctrl = None
 
@@ -167,19 +176,28 @@ def parse_args():
 def run():
     global ctrl
 
+    Trace.mark('run.enter')
     args = parse_args()
+    Trace.mark('args.parsed')
 
     # Set signal handler
     signal.signal(signal.SIGTERM, on_exit)
 
     # Create ioloop
     ioloop = tornado.ioloop.IOLoop.current()
+    Trace.mark('ioloop.created')
 
     # Set ObjGraph signal handler
     if args.debug: Debugger(ioloop, args.debug)
 
     # Start server
-    web = Web(args, ioloop)
+    with Trace.span('web.init'):
+        web = Web(args, ioloop)
+    Trace.mark('listen', port=args.port, addr=args.addr)
+
+    # Notify systemd we are ready (no-op when not under systemd).
+    Trace.sd_notify('READY=1\nSTATUS=listening on %s:%d\n' %
+                    (args.addr, args.port))
 
     try:
         ioloop.start()