Remove BadBlockManager from ProtocolSpec

[siladu]

This is a bug because when you transition to a new fork schedule, you lose your existing bad block cache.

Could do with knowing why it was added there in the first place before we blindly remove it though.

[siladu]

Probably not a bug so much as a feature with unintended consequences.

It make sense to partition bad blocks by fork since they blocks may be valid/invalid depending on different fork rules.

It does mean that the previous fork's bad blocks are unavailable in the debug_getBadBlocks endpoint, however that's a very minor edge case, probably not worth the effort so closing this.

If we implemented the "archive" cache mentioned in #5460 and made it agnostic of fork/protocol schedule then that may fix this issue as well.