validation: add missing insert to m_dirty_blockindex

[mzumsande]

When the status of a block index is changed, we must add it to m_dirty_blockindex or the change might not get persisted to disk.
This is missing from one spot in FindMostWorkChain(), where BLOCK_FAILED_CHILD is set.
Since we have code that later sets missing BLOCK_FAILED_CHILD during the next startup, I don't think that this can lead to bad block indexes in practice, but I still think it's worth fixing.

[DrahtBot]

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

Reviews

See the guideline for information on the review process.

Type	Reviewers
ACK	TheCharlatan, stickies-v

If your review is incorrectly listed, please react with 👎 to this comment and the bot will ignore it on the next update.

[TheCharlatan]

ACK e639364

[stickies-v]

ACK e639364

Couldn't find any other instances of CBlockIndex::nStatus updates not being added to m_dirty_blockindex. As pointed out by mzumsande, I think indeed this should be corrected nicely upon startup since LoadBlockIndex() iterates from start->end, as opposed to the reverse end->start being done here.

[Sjors]

Should we make nStatus private and only allow writing to it through a helper method that marks the index entry dirty?

[mzumsande]

Should we make nStatus private and only allow writing to it through a helper method that marks the index entry dirty?

Would be nice if we didn't have to remember, I think there would be a few complications:

• CBlockIndex is an object that can exist and be updateable independent from whether it's part of our block index database - I could imagine there are spots that update its members that don't require marking it dirty.
• Do we only want to mark nStatus private, and leave all other members public? If it's the latter, then it could become quite a big change...