Skip db name parsing if mts_accessed_dbs == 254 (julien-duponchelle#369)

I interpreted `mts_accessed_dbs` as db count in PR#360 (released in 0.27).
However after reproducing kvitek's exception in Issue#278,
I did some more research how it is set:
https://github.com/mysql/mysql-server/blob/6846e6b2f72931991cc9fd589dc9946ea2ab58c9/sql/log_event.cc#L3598-L3602
(log_event.h /log_event.cc is responsible for writing low-level binlog,
according to
https://dev.mysql.com/doc/internals/en/source-files-related-to-the-binary-log.html)

Turns out that if the value of the variables equals 254,
the variable is no longer interpreted as the db count,
but represent a following case:
1. db count equals 1
2. The name of the only db is blank.

Therefore, I edited event.py so that after the value 254 is encountered,
no further parsing is done regarding db names.

So far I have reproduced the issue for `BeginEvent`.
However, I have not yet figured out to incorporate into unittest.
I will add tests ASAP.

Author: dongwook-chan

pymysqlreplication/event.py

@@ -257,6 +257,17 @@ def _read_status_vars_value_for_key(self, key):
                 self.host = self.packet.read(host_len)
         elif key == Q_UPDATED_DB_NAMES:               # 0x0C
             mts_accessed_dbs = self.packet.read_uint8()
+            """
+            mts_accessed_dbs < 254:
+                `mts_accessed_dbs` is equal to the number of dbs
+                acessed by the query event.
+            mts_accessed_dbs == 254:
+                This is the case where the number of dbs accessed
+                is 1 and the name of the only db is ""
+                Since no further parsing required(empty name), return.
+            """
+            if mts_accessed_dbs == 254:
+                return
             dbs = []
             for i in range(mts_accessed_dbs):
                 db = self.packet.read_string()