#3223 follow-up

[sistemd]

Just two things stood out:

1. The &HashMap<u64, (u32, ConsensusFinalizedL2Block)> is -imo- non-obvious, and is passed around quite a bit. I'd document it somehow or maybe alias the numeric types in a way where its intent becomes clearer.

2. The fake_decided_blocks trick might deserve an issue. Not sure. But it looks hacky and maybe it's a case that validate_block_info should be ready to support.

Originally posted by @t00ts in #3223 (review)