ci: more printing about state of master (#9058)

Two problems combined into one:
1. the session that `_wait_for_master` used had the system's default retry policy (which takes about 15 seconds before it gives up)
2. `_wait_for_master` reported every 60 iterations through its loop (which itself was limited by the number of iterations) rather than every 60 seconds. This meant that it only actually reported once every 16 minutes.

CircleCI had been killing workflows that didn't print text after 10 minutes. So when migrations took more than 10 minutes, the whole test would fail.

This PR fixes both of problems (1) and (2).

.circleci/scripts/wait_for_perf_migration_upload_results.py

@@ -16,19 +16,23 @@
 def _wait_for_master() -> None:
     print("Checking for master")
     cert = certs.Cert(noverify=True)
-    sess = api.UnauthSession("http://127.0.0.1:8080", cert)
+    sess = api.UnauthSession("http://127.0.0.1:8080", cert).with_retry(0)
 
     # 2 hours is the most a migration can take, with this setup.
     # If a migration takes longer than that we have hit an issue a customer will likely hit too.
-    for i in range(2 * 60 * 60):
+    time_start = time.time()
+    time_last_report = time_start
+    while time_start - time.time() < 2 * 60 * 60:
         try:
             r = sess.get("info")
             if r.status_code == requests.codes.ok:
+                print(f"Master up and available after {int(time.time() - time_start) / 60} minutes")
                 return
         except api.errors.MasterNotFoundException:
             pass
-        if i % 60 == 0:
-            print("Waiting for master to be available...")
+        if time.time() - time_last_report > 60:
+            time_last_report = time.time()
+            print(f"Waiting for master ({int(time_last_report - time_start) / 60} minutes elapsed)")
         time.sleep(1)
     raise ConnectionError("Timed out connecting to Master")