Use http Keep-Alive, else BigQuery queries are ~seconds slower than necessary

- Before (without Keep-Alive): ~3-7s for BigQuery `select 3` with an already cached result
- After (with Keep-Alive): ~1.5-3s
- Query sends these 6 http requests and runtime appears to be dominated by network RTT:
```
http: method[POST], url[https://www.googleapis.com/bigquery/v2/projects/foo-bar/jobs/], body[{"kind": "bigquery#job", "configuration": {"priority": "INTERACTIVE", "query": {"allowLargeResults": false, "useLegacySql": false, "useQueryCache": true, "query": "select 3", "userDefinedFunctionResources": []}, "dryRun": false}}]
http: method[GET], url[https://www.googleapis.com/bigquery/v2/projects/foo-bar/queries/job_sIi1HRqfyHkRB5_MkfWFjtZ60XM?startIndex=0&timeoutMs=30000&maxResults=0], body[None]
http: method[GET], url[https://www.googleapis.com/bigquery/v2/projects/foo-bar/jobs/job_sIi1HRqfyHkRB5_MkfWFjtZ60XM], body[None]
http: method[GET], url[https://www.googleapis.com/bigquery/v2/projects/foo-bar/datasets/_2f96775300d8858559d2bd23c05bad0392345e30/tables/anon921947a4e6645dc2b34411c365f9a45e0895d5a4], body[None]
http: method[GET], url[https://www.googleapis.com/bigquery/v2/projects/foo-bar/datasets/_2f96775300d8858559d2bd23c05bad0392345e30/tables/anon921947a4e6645dc2b34411c365f9a45e0895d5a4], body[None]
http: method[GET], url[https://www.googleapis.com/bigquery/v2/projects/foo-bar/datasets/_2f96775300d8858559d2bd23c05bad0392345e30/tables/anon921947a4e6645dc2b34411c365f9a45e0895d5a4/data?maxResults=25], body[None]
```

Author: jdanbrown

datalab/utils/_http.py

@@ -53,6 +53,16 @@ class Http(object):
   """A helper class for making HTTP requests.
   """
 
+  # Reuse one Http object across requests to take advantage of Keep-Alive, e.g.
+  # for BigQuery queries that requires at least ~5 sequential http requests.
+  #
+  # TODO(nikhilko):
+  # SSL cert validation seemingly fails, and workarounds are not amenable
+  # to implementing in library code. So configure the Http object to skip
+  # doing so, in the interim.
+  http = httplib2.Http()
+  http.disable_ssl_certificate_validation = True
+
   def __init__(self):
     pass
 
@@ -109,15 +119,8 @@ def request(url, args=None, data=None, headers=None, method=None,
     if method is None:
       method = 'GET'
 
-    # Create an Http object to issue requests. Associate the credentials
-    # with it if specified to perform authorization.
-    #
-    # TODO(nikhilko):
-    # SSL cert validation seemingly fails, and workarounds are not amenable
-    # to implementing in library code. So configure the Http object to skip
-    # doing so, in the interim.
-    http = httplib2.Http()
-    http.disable_ssl_certificate_validation = True
+    # Authorize with credentials if given.
+    http = Http.http
     if credentials is not None:
       http = credentials.authorize(http)
     if stats is not None: