Send store.finalized block hash to EL

[mkalinin]

Explicitly states that store's finalized checkpoint must be used to obtain finalized_block_hash that is sent to EL. Apparently, all CL clients are already following this logic.

@hwwhww we would need to create tests checking that store.finalized used instead of head_state.finalized to obtain finalized_block_hash. These values can't diverge according to current specification, as tips with finalized checkpoint not matching to the store's one are filtered out and not eligible for the head. Not sure if we can implement such tests without changing the fork choice spec.

[hwwhww]

@mkalinin

we would need to create tests checking that store.finalized used instead of head_state.finalized to obtain finalized_block_hash. These values can't diverge according to current specification, as tips with finalized checkpoint not matching to the store's one are filtered out and not eligible for the head. Not sure if we can implement such tests without changing the fork choice spec.

To provide test vectors, we need a new test format to mock notify_forkchoice_updated. I think it may be something like #2639 proposed, but probably in a simpler form. What do you think?

---

I pushed basic unit tests (270d930) to verify code logic of this PR.

[mkalinin]

To provide test vectors, we need a new test format to mock notify_forkchoice_updated. I think it may be something like #2639 proposed, but probably in a simpler form. What do you think?

Do you think that by mocking notify_forkchoice_updated we can bypass the requirement of store.finalized and head_state.finalized to diverge? Or you think this requirement is not needed to be satisfied to write tests covering proposed change?