Add oracle call for retrieving the public key associated to a contract

[spalladino]

Today contracts use a mix of addresses and public keys for referring to users in Noir contracts. For instance, the zk_token requires the pubkeys of sender and recipient (for decrypting/encrypting notes) but not their addresses, and its transfer function does not check that the sender is actually the msg.sender (meaning that a leaked encryption key causes loss of assets, even if the signing key is not compromised). And for the non_native token used in cross-chain tests, we're needing both the address and the pubkey as arguments.

Now, with the ongoing changes for supporting Account Abstraction, we're calculating deployment addresses as the hash of a public key, a salt, contract code, and constructor args. We can tweak this to hash(deployer_public_key, h1), where h1 := hash(salt, function_tree_root, constructor_hash), so that given public_key and h1, we can verify the resulting address.

This means we could implement an oracle call that, given an address, returns the corresponding public_key and h1, and Noir can verify that the oracle response is correct by hashing them and checking that the result is equal to the address (which Joe had originally suggested for getting the pubkey in the entrypoint of an account contract). And we could use this to stop passing public keys around in contracts, and just use addresses for referring to a user within contracts. This will require a method for sharing public_keys and h1s though (pretty much what Vitalik had identified on the Three Transitions post, but that's a longer story), but we can probably work around that for the devnet.

This'd mean we do not abstract encryption and nullifying keys for now, but if we hide this get_public_key logic behind a Noir library, it should be easy to change in the near future.