client: Expose future based submissions

Currently all the KCIDB submission APIs are synchronous, meaning we have
network latency in the call which isn't ideal for large numbers of small
submissions. The underlying BigQuery PubSub APIs offer an API based on
futures which should hopefully allow us to do better here, allowing the
client to submit data and then later check that it has actually been sent
so that the PubSub library can batch I/O or do it asynchronously.

This is already wired up at the publisher level so just add a top level
client method which exposes the future based API. We may want to also
expose something of the flow control parameters via the PublishFlowControl
passed when creating the client, though the current defaults (1000 messages
and 10Mb in flight) look reasonable.

Signed-off-by: Mark Brown <broonie@kernel.org>

kcidb/__init__.py

@@ -83,6 +83,30 @@ def submit(self, data):
             raise NotImplementedError
         return self.mq_publisher.publish(data)
 
+    def submit_future(self, data):
+        """
+        Submit reports without blocking for the interaction with the
+        database.
+
+        Args:
+            data:   A JSON object with the report data to submit.
+                    Must adhere to the current, or an earlier version of I/O
+                    schema. Note that this function will not validate the
+                    submitted data.
+
+        Returns:
+            A future which will return the Submission ID string.
+
+        Raises:
+            `NotImplementedError`, if not supplied with a project ID or an MQ
+            topic name at initialization time.
+        """
+        assert io.SCHEMA.is_compatible(data)
+        assert LIGHT_ASSERTS or io.SCHEMA.is_valid(data)
+        if not self.mq_publisher:
+            raise NotImplementedError
+        return self.mq_publisher.publish_future(data)
+
     def submit_iter(self, data_iter, done_cb=None):
         """
         Submit reports returned by an iterator.