[MeABzZz]

What this does

Test-only. Ports tests/connect_pow_check.rs from #23 onto current main. Main fixed the bug in bbdaccb: connect_block now checks proof of work unconditionally. This PR only pins that behaviour, and there are no src/ changes.

The vectors use mainnet block 1 (00000000839a8e68…eb6048), rebuilt from its coinbase so the merkle root is checked as well.

test	what it pins
vector_is_mainnet_block_1	The vector hashes to the real block-1 hash and meets its own target
connect_accepts_mainnet_block_1	A valid block connects
connect_rejects_hash_above_target_with_correct_bits	Correct nBits with a nonce that misses the target is rejected ("proof of work")
connect_rejects_hash_above_target_without_ancestor_lookup	New vs #23. The same rejection with no difficulty_ancestor lookup. Before bbdaccb this path skipped PoW.
connect_regtest_pow_boundary	Regtest boundary: a hash at the target passes and one above it fails

Verification

• On main (a9fb32d, which includes bbdaccb), cargo test --test connect_pow_check gives 5 passed.
• Mutation check: replacing the check at src/block/connect.rs:3313 (if !crate::pow::check_proof_of_work(..)) with if false gives 3 failed / 2 passed. The 3 rejection tests fail and the 2 acceptance and vector tests still pass, as expected. The file was then restored.
• rustfmt --edition 2024 --check on the new file is clean.
• Clippy on this test target reports only useless_conversion on script_sig/script_pubkey .into(). That matches the existing test style (segwit_suite.rs, integer_overflow_edge_cases.rs, …), so I left it as is. The whole-crate --all-targets clippy errors on main are pre-existing and unrelated to this PR.

Review notes

• Builds on: bbdaccb (unconditional PoW check in connect_block).
• Merge order: independent and can merge any time.
• Supersedes: fix(consensus): reject a block whose hash is above its target on connect #23. After this merges, fix(consensus): reject a block whose hash is above its target on connect #23 (whose fix main already has via bbdaccb) can be closed. I haven't closed it; that's the maintainer's call.