Model for keystore, key, subkey interactions

[ronaldtse]

Description

From @ni4 in #497 :

In my opinion subkey should be part of the key, not the separate entity of the keystore. So, for instance, for decryption/verification operation it is searched by keyid, and for encryption/signing - by user id of the main key. @dewyatt, @flowher, @ronaldtse what do you think?

[ronaldtse]

In terms of structure I think the subkey concept should be just like any other key, but is simply linked to a parent key. However, when we look for a key we should first find the parent key and then the subkey.

[ni4]

Sure, subkey should be treated as the key in terms of 'key material' in all cryptography operations and primitives. However, when we get to the trust check, identity check and so on we should treat it as the part of the main key. Technically we may implement it in the different ways, but we must have links from the subkey to main key and back - from the main key to all the subkeys.

[ronaldtse]

@ni4 agree that an association both ways is necessary between the parent and sub-keys.

[ronaldtse]

@ni4 Is this something we need to follow up on?

[ni4]

@ronaldtse Yep, I think this is still on the table.
@dewyatt what do you think about changing the current behavior where main keys and subkeys are separate entities in the keyrings?

[dewyatt]

@ni4 Can you elaborate in some more detail (maybe a mockup or something)?