Default value of the Cluster.replaced_deployment_info field was changed without appropriate migration

The default value of the Cluster.replaced_deployment_info field was changed from '{}' to '[]' without an appropriate DB migration script during 9.0 release [1] and [2]. Also this field was defined as MutableList and this type does to accept dict values. The problem appears after migration of the database in case when there is at least one cluster created in previous releases < 9.0 that does not have replaced deployment information.

Steps to reproduce:
1. Create environment using the 8.0 release
2. Upgrade master node to the 9.0 release (involve applying of migration scripts to synchronize the database to 9.0)
3. Try to get a list of environments using the fuel client:
    fuel environment

Expected results:
A list of existing environments will be printed.

Actual result:
This command fails with the error:

    500 Server Error: Internal Server Error for url: http://10.20.0.2:8000/api/v1/clusters/ (Attribute 'replaced_deployment_info' does not accept objects of type <type 'dict'>)

The traceback from /var/log/nailgun/app.log:

    2016-04-08 19:01:49.225 ERROR [7f2b8f602880] (base) Unexpected exception occured
    Traceback (most recent call last):
      File "/usr/lib/python2.7/site-packages/nailgun/api/v1/handlers/base.py", line 285, in content_json
        resp = func(cls, *args, **kwargs)
      File "/usr/lib/python2.7/site-packages/nailgun/api/v1/handlers/base.py", line 446, in GET
        return self.collection.to_json(q)
      File "/usr/lib/python2.7/site-packages/nailgun/objects/base.py", line 437, in to_json
        iterable=iterable
      File "/usr/lib/python2.7/site-packages/nailgun/objects/base.py", line 421, in to_list
        use_iterable
      File "/usr/lib64/python2.7/site-packages/sqlalchemy/orm/loading.py", line 86, in instances
        util.raise_from_cause(err)
      File "/usr/lib64/python2.7/site-packages/sqlalchemy/util/compat.py", line 200, in raise_from_cause
        reraise(type(exception), exception, tb=exc_tb, cause=cause)
      File "/usr/lib64/python2.7/site-packages/sqlalchemy/orm/loading.py", line 71, in instances
        rows = [proc(row) for row in fetch]
      File "/usr/lib64/python2.7/site-packages/sqlalchemy/orm/loading.py", line 432, in _instance
        state.manager.dispatch.load(state, context)
      File "/usr/lib64/python2.7/site-packages/sqlalchemy/event/attr.py", line 256, in __call__
        fn(*args, **kw)
      File "/usr/lib64/python2.7/site-packages/sqlalchemy/ext/mutable.py", line 451, in load
        val = cls.coerce(key, val)
      File "/usr/lib/python2.7/site-packages/nailgun/db/sqlalchemy/models/mutable.py", line 141, in coerce
        return MutableBase.coerce(key, value)
      File "/usr/lib64/python2.7/site-packages/sqlalchemy/ext/mutable.py", line 403, in coerce
        raise ValueError(msg % (key, type(value)))
    ValueError: Attribute 'replaced_deployment_info' does not accept objects of type <type 'dict'>

Reproducibility:
This bug reproduces constantly if there is at least one environment that was created in the previous releases (<9.0) without the replaced deployment information.

Workaround:
Database entries can be fixed by the next SQL command:
    UPDATE clusters SET replaced_deployment_info = '[]' WHERE replaced_deployment_info = '{}';

Impact:
This bug is a blocker for the upgrade procedure of the master node from 8.0 to 9.0.

Links:
[1] The default value was changed to '[]' and the type was changed to MutableList in the next change request: https://review.openstack.org/#/c/249723/
[2] This change was made as a part of this bug: https://bugs.launchpad.net/fuel/+bug/1482658