Load assets lazily. Part of #401.

Changes

• Add {AssetVault, PartialVault}::open.
• Introduces AssetWitness as a type returned by AssetVault::open and for use in the DataStore interface.
  • This results in a type safe API and makes the API more consistent with AccountTree and NullifierTree.
• Changes TransactionExecutorHost to require a STORE: DataStore rather than STORE: MastForestStore. This is needed so the host can request data from the store.
• React to before add/remove asset events by checking if the merkle store already contains a merkle path for this asset and:
  • if so, continue
  • if not, request it from the data store.
• Tracks the native_account_header in the base host to be able to access the initial vault root of the account. With lazy storage map loading we'd also need the account storage header for similar reasons.
• Adds TransactionContextBuilder::enable_partial_loading to selectively "enable" lazy loading for testing purposes. Enabling it immediately would break 50+ tests and not all of them are easily fixable (see below).
• Removes the advice_inputs_from_transaction_witness_are_sufficient_to_reexecute_transaction test, which has frequently been difficult to maintain. Here also, due to the host requiring a DataStore rather than just a MastForestStore impl it was difficult to keep up to date. I think what this test asserts is essentially implicitly covered by the fact that we can prove, which doesn't have access to a data store. It would definitely be nice to keep the test, but it is difficult to maintain.
• Add mock::util library for creating notes more easily. This can be extended with many more test functionality in the future, such as exec-utable wrappers around mock account APIs, which is currently annoying to use, due to the call convention.

Note that the scope of this PR is asset witness for the native account. For now I'm assuming we only request native account witnesses in the code, but this would eventually be changed to handle foreign accounts as well.

Follow-Ups:

• Input Vault currently requests missing assets one by one, causing many individual data store lookups. These should be requested before the transaction and in bulk. We could probably do it in this PR, but I just left it like that for now as other things were more important to get done first.
• Move TransactionKernelError to miden-tx. It's defined in miden-lib but really should be in miden-tx (see DataStoreError issue in the code).
• PartialVault::add should take AssetWitness instead. This is a bit too much to do to include in this PR so can be a follow-up.
• We should consider whether we still need to expose AssetVault::asset_tree. Instead, I think we should replace its usages with AssetVault::open or add missing, type-safer alternatives to AssetVault directly. It's still fine to be able to access the underlying SMT, but generally, there should be a better API for this. Again, makes it type safer and more in line with AccountTree and NullifierTree.
• Is there a way to convert a Word into the Felt that represents an SMT's LeafIndex? If not, should there be something like that in miden-crypto? It'd be much better than having vault_key[3] in places which is basically a magic constant. Basically, making Smt::key_to_leaf_index publicly available would be great.
• To use TransactionContext::execute_code with lazy loading, we'd have to make MockHost also request missing witnesses from a store. It currently doesn't have a store at all. I'm not sure the mock host makes sense any longer if we have to add more logic to it that duplicates the logic in the tx host. I think we should check if we can replace MockHost with TransactionExecutorHost for CodeExecutor. Otherwise we'll have to duplicate the lazy loading things.
  • I tried this but one prerequisite for this is Allow access to FastProcessor memory after execution miden-vm#2052, because we rely on accessing Process memory after execute_code in a few places.
• Should we have a convenience method on SmtProof for this?

let smt_proof = SmtProof::from(asset_witness);
let authenticated_nodes = smt_proof
.path()
.authenticated_nodes(smt_proof.leaf().index().value(), smt_proof.leaf().hash())
.expect("leaf index is u64 and should be less than 2^SMT_DEPTH");