Operational inconsistencies with new storage_cache role

I noticed two strange things
1. storage_cache never appears as a role in status json.

For example:
```
[nav] In [39]: sorted([(v['address'], [x['role'] for x in v['roles']]) for k, v in json.loads(db[b'\xff\xff/status/json'])['cluster']['processes'].items()])
Out[39]:
[('127.0.0.1:4000', ['coordinator', 'proxy', 'storage']),
('127.0.0.1:4001', ['master', 'storage']),
('127.0.0.1:4002', ['data_distributor', 'storage']),
('127.0.0.1:4003', ['storage', 'resolver']),
('127.0.0.1:4004', ['cluster_controller', 'storage']),
('127.0.0.1:4005', ['ratekeeper', 'storage']),
('127.0.0.1:4006', ['proxy', 'storage']),
('127.0.0.1:4007', ['log', 'storage']),
('127.0.0.1:4008', ['log', 'storage']),
('127.0.0.1:4009', ['storage']),
('127.0.0.1:4010', ['proxy', 'storage']),
('127.0.0.1:4011', ['log', 'storage']),
('127.0.0.1:4012', [])]
```
127.0.0.1:4012 is started with `--class storage_cache`, but it doesn't show up in status json.

2. Using the `setclass` fdbcli command with class storage_cache doesn't seem to have an effect.