In-memory encryption of privkeys

[fluffypony]

To prevent them ever being paged to disk unencrypted, they should be encrypted in-memory until they are needed.

Reference in terms of how Bitcoin does it (we don't have to follow their model 1:1) - https://bitcointalk.org/index.php?topic=8728.0

per @gmaxwell - "best effort mlocked, zeroized when the wallet is no longer unlocked or when freed."

[ghost]

@fluffypony has anyone taken this forward?

[fluffypony]

@nanoakron no not yet

[dEBRUYNE-1]

+enhancement

[moneromooo-monero]

Partially done with jroelofs' patches above, still some left to do though.

[stoffu]

IIUIC the ringct code lumps the different kinds of data (pubkey/seckey/hash) into a single type rct::key, and thus the encryption for the secret key isn’t achieved yet.