lib/scylla_cloud.py: change retry logic of curl() to exponential backoff

Since IaaS services recommended to use exponential backoff logic when
retries to call metadata services, we should do it in our scripts.

Referenced implementation in https://developers.google.com/analytics/devguides/reporting/core/v3/errors?hl=en

see https://docs.aws.amazon.com/general/latest/gr/api-retries.html

Related with scylladb/scylladb#13442

lib/scylla_cloud.py

@@ -18,6 +18,7 @@
 import distro
 import base64
 import datetime
+import random
 from subprocess import run, CalledProcessError
 from abc import ABCMeta, abstractmethod
 
@@ -69,8 +70,10 @@ def curl(url, headers=None, method=None, byte=False, timeout=3, max_retries=5, r
                     return res.read()
                 else:
                     return res.read().decode('utf-8')
-        except (urllib.error.URLError, socket.timeout):
-            time.sleep(retry_interval)
+        except (urllib.error.URLError, socket.timeout) as e:
+            exp_retry_interval = (retry_interval ** retries) + random.random()
+            logging.error(f'Failed to access {url}: {e.reason}, retrying in {exp_retry_interval}')
+            time.sleep(exp_retry_interval)
             retries += 1
             if retries >= max_retries:
                 raise