refactor: extract stateless service

[sergey-melnychuk]

Get rid of syncing concept - it makes no sense completely in a stateless setup. Instead each request must be resolved to the currently latest block (queried from L1 or L2, according to the selected trust mode: #657). Then all processing and proof validation must happen respectful of this current block.

Using properly extracted Ethereum (#668) and Starknet (#669) clients, such implementation should be rather straightforward:

1. Lookup if there is a cached latest block that is still fresh enough (timeout).
2. If not: lookup latest block & state root according to the selected trust mode.
3. Make the updated block & state root available for state proofs validation.