mining: move BlockAssembler statics into BlockTemplateManager #36357

Based on #36356. Together with it, this completes the "Remove remaining global states" TODO in #33758.

getmininginfo reports currentblockweight and currentblocktx from two process-wide statics, BlockAssembler::m_last_block_weight and m_last_block_num_txs, which every BlockAssembler::CreateNewBlock() call overwrites.

Like the getblocktemplate statics, they are shared by every node context in the process and nothing checks their locking. They also record templates that were not built for mining: the mempool fee estimator builds its own templates with BlockAssembler, so getmininginfo can report the fee estimator's template instead of the last mining template.

This PR:

• makes the two fields per-BlockAssembler instance, still set at the same point in CreateNewBlock();
• has BlockTemplateManager::CreateNewTemplate() record them in a member guarded by cs_main, which getmininginfo reads. CreateNewTemplate() now holds cs_main across the build, as CreateNewBlock() already did internally, so concurrent builds cannot record their stats out of order;
• adds a unit test showing that manager builds are recorded, that the latest one wins, and that a direct BlockAssembler build is ignored.

In short: the last global mining state moves out of BlockAssembler, and getmininginfo reports the last template built for mining rather than whatever the node assembled last.

<!--e57a25ab6845829454e8d69fc972939a-->

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

<!--006a51241073e994b41acfe9ec718e94-->

Code Coverage & Benchmarks

For details see: https://corecheck.dev/bitcoin/bitcoin/pulls/36357.

<!--021abf342d371248e50ceaed478a90ca-->

Reviews

See the guideline and AI policy for information on the review process.

If your review is incorrectly listed, please copy-paste <code>&lt;!--meta-tag:bot-skip--&gt;</code> into the comment that the bot should ignore.

<!--174a7506f384e20aa4161008e828411d-->

Conflicts

Reviewers, this pull request conflicts with the following ones:

• #36525 _{<img src="https://drahtbot.space/ack_count/bitcoin/bitcoin/36525.svg">} (mining: add getTotalFees(), protect m_block_template with mutex by Sjors)
• #36451 _{<img src="https://drahtbot.space/ack_count/bitcoin/bitcoin/36451.svg">} (refactor: Use NotNull for CreateNewBlock and pimpls by maflcko)
• #35671 _{<img src="https://drahtbot.space/ack_count/bitcoin/bitcoin/35671.svg">} (mining: add TxCollection to bandwidth-efficiently validate external block templates by Sjors)
• #35569 _{<img src="https://drahtbot.space/ack_count/bitcoin/bitcoin/35569.svg">} (Encapsulation for CTransaction by purpleKarrot)
• #35511 _{<img src="https://drahtbot.space/ack_count/bitcoin/bitcoin/35511.svg">} (RFC: consensus: Make CAmount a class by hodlinator)
• #33922 _{<img src="https://drahtbot.space/ack_count/bitcoin/bitcoin/33922.svg">} (mining: add getMemoryLoad() and track template non-mempool memory footprint by Sjors)

If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.

<!--5faf32d7da4f0f540f40219e4f7537a3-->

Concept ACK

Now that #34075 is merged, we can also move its cache in there, i.e cache the block template based on time and tip changes. This brings us multiple advantages

i) We consolidate the block template caching for the RPC and fees to the block template manager for better code organisation and reuse. ii) The mempool fee rate estimator's reliance on the chainstate manager is reduced; the only dependence after that will be on the chain tip, which can also be removed by using chain tip change validation signals. After that, the mempool fee rate estimator will be independent of the mempool and chainstate manager, in replacement with block template manager.