CLI API for dat schemas

[max-mapper]

inspired by a interaction with @ekg today in IRC:

2:56 PM <erikg> dat init; (env)% echo '{"hello": "world"}' | dat import --json
2:56 PM <erikg> % dat cat
2:56 PM <erikg> {"hello":"world","key":"ci0ps81a20000uia4ysz8957b","version":1}
2:57 PM <erikg> % echo '{"hellox":"worldz","key":"ci0ps81a20000uia4ysz8957b","version":2}' | dat import --json
2:57 PM <erikg> % dat cat
2:57 PM <erikg> {"hello":null,"hellox":"worldz","key":"ci0ps81a20000uia4ysz8957b","version":2}
2:57 PM <erikg> what i don't understand is why 'hello' is retained in the object

The reason for this behavior is that dat uses a schema to store data (using Protocol Buffers under the hood). When you write JSON it automatically adds any new keys to the schema. This is a pretty auto-magic process right now. To help demystify this we should add CLI APIs to view and edit the schema.

Strawman proposal:

• dat schema would print a JSON version of the current protobuf schema
• dat schema edit would open the schema in the default editor (similar to typing git commit)
• also possibly dat schema import foo.json to let you supply your own schema from a file. Not sure the exact semantics of this one
• also maybe we should by default write a message to stderr whenever the schema gets changed, just to make it more visible to the user that there is a schema under the hood