Fix handling of four byte utf8 manifests

- Adds the charset: utf-8 to all the manifest responses
- Makes sure we connect to MySQL in utf8mb4 mode, to ensure we can properly read and write 4-byte utf8 strings
- Adds tests for all of the above

data/database.py

@@ -70,6 +70,12 @@ SCHEME_RANDOM_FUNCTION = {
 }
 
 
+_EXTRA_ARGS = {
+  'mysql': dict(charset='utf8mb4'),
+  'mysql+pymysql': dict(charset='utf8mb4'),
+}
+
+
 def pipes_concat(arg1, arg2, *extra_args):
   """ Concat function for sqlite, since it doesn't support fn.Concat.
       Concatenates clauses with || characters.
@@ -315,6 +321,10 @@ def _db_from_url(url, db_kwargs, connect_timeout=DEFAULT_DB_CONNECT_TIMEOUT,
     db_kwargs.pop('stale_timeout', None)
     db_kwargs.pop('max_connections', None)
 
+  for key, value in _EXTRA_ARGS.get(parsed_url.drivername, {}).iteritems():
+    if key not in db_kwargs:
+      db_kwargs[key] = value
+
   if allow_retry:
     driver = _wrap_for_retry(driver)