block_builder tests: fix and use new infrastructure

[plafer]

#77 introduced a new infrastructure useful especially for the block_builder tests. It is currently incomplete, and needs to be fixed before every block_builder test can be updated to use it. Specifically

• MockStoreSuccessBuilder and build_expected_block() need to compute the correct notes root
• MockStoreSuccessBuilder: new() needs to be replaced with a few different constructors (including from_batches()) which allow it to know the initial account IDs and state.

Once this is done, use this new test code everywhere, and remove the obsolete DummyProvenTxGenerator.

[hackaugusto]

@plafer could you add more context on what needs to be done? What should be achieved by this issue?

[plafer]

The ultimate goal is to get rid of DummyProvenTxGenerator.

The objects that I started creating to replace/enhance its functionality are:

• MockProvenTxBuilder: has methods to create a mock ProvenTransaction
• TransactionBatchConstructor: provides convenience functions to create batches
• MockBlockBuilder: a builder to create a mock block

These are already used in some tests (except for TransactionBatchConstructor) - all other tests should be written using those. You'll probably need to adapt them too. Feel free to redesign them too if you prefer - again, as long as we get rid of DummyProvenTxGenerator.

[hackaugusto]

Thanks for the update.

The ultimate goal is to get rid of DummyProvenTxGenerator.

And why do we want to get rid of it? What features is it lacking that are introduced by the other mocks?

[plafer]

It was a hack at the time before we had StarkProof::new_dummy(). The only way to create a StarkProof object was by running the winterfell prover on a bogus problem (i.e. all this). Running this makes tests slower, and just isn't clean.

With the other mocks, I also took this change as an opportunity to create objects that make tests more readable and reduce the amount of copy/paste. It's still a WIP though (and hence this issue).