Python client lockstep: per-topology difficulty + mineable-topology whitelist

[rcarback]

Draft — do not merge until the runtime change is deployed. Depends on quip-protocol-rs MR !42 (per-topology difficulty + mineable-topology whitelist). The new runtime API/encoding this targets does not exist on chain until that ships.

Summary

Keeps the Python client in lockstep with the Substrate change that binds PoW difficulty per topology and gates mining on a root-controlled whitelist. Also retires the per-solution participation System.remark in favor of the pallet's native miner registry.

Changes — per-topology difficulty + whitelist

• substrate/client.py
  • query_difficulty(topology_hash=None) now reads the per-topology QuantumPow.Difficulties map (resolves DefaultTopology when None; absent entry returns None).
  • New query_difficulty_for(topology_hash) (runtime API QuantumPowApi_difficulty_for, decoding Option<DifficultyConfig>).
  • New query_mineable_topologies() (runtime API QuantumPowApi_mineable_topologies, decoding Vec<H256>).
  • New add_mineable_topology / remove_mineable_topology submit helpers.
• substrate/scale_codec.py: _decode_h256_vec (compact length + N×32 bytes).
• substrate/miner_bootstrap.py: _maybe_seed_chain registers the topology before seeding difficulty and passes topology_hash to set_difficulty (now a two-argument call).
• tools/register_advantage2.py: restructured to compute the target hash, register, whitelist (add_mineable_topology), set_difficulty(topology_hash, …), then set_default_topology — the order the new runtime requires (set_difficulty needs the topology registered; set_default_topology rejects non-whitelisted targets). This file was previously an uncommitted local tool and is now tracked.
• tools/download_and_validate_wins.py: no change needed — already re-validates against the per-topology snapshot difficulty.

Changes — drop participation System.remark for pallet-native tracking

The per-solution participation remark was a stopgap from before the pallet had a miner registry: a free-form System.remark JSON blob per solution to record that a node was mining. The pallet now tracks this natively — register_miner reserves a deposit and records identity, and submit_proof increments proofs_submitted/proofs_won in the Miners map — so the remark is redundant transport for data the chain already owns. register_miner is already submitted at bootstrap (_ensure_registered).

Removed the whole producer→consumer chain:

• shared/base_miner.py: participating_cb param + dispatch-gate emit + _participation_extra.
• QPU/dwave_miner.py: _participation_extra override (budget_seconds).
• shared/miner_worker.py: _emit_participating and its wiring.
• substrate/miner_controller.py: _mark_participating, _submit_participation_remark, the op=="participating" drain branch, the participation dedup state, the retry constants, and the now-unused submit_remark import.
• Tests covering the removed path in test_base_miner_pump.py, test_dwave_stream_stop.py, test_substrate_miner_controller.py.

The descriptor/identify hardware-specs remarks keep using System.remark — register_miner takes no payload, so there is no pallet equivalent for broadcasting GPU/CPU specs.

Testing

• ruff check / ruff format --check clean on touched files.
• Unit and mocked tests pass (scale-codec helpers, pool client, mining-snapshot decode, miner bootstrap). New _decode_h256_vec tests build real SCALE wire bytes; new runtime-API helpers covered via monkeypatched _state_call.
• Participation removal: 80 passed across the three affected test files; no new failures.
• Live-chain integration tests are skipif-gated on a running node and require the deployed runtime to exercise end to end.

Live test gating

test_controller_submits_proof_end_to_end and test_controller_long_haul_multi_block drive _maybe_seed_chain, which reads/writes difficulty through the per-topology QuantumPow.Difficulties map. They now probe the connected chain's metadata (_chain_has_per_topology_difficulty) and skip cleanly against a pre-MR!42 runtime instead of hard-failing — mirroring the existing _chain_requires_hybrid_signer gate. They exercise end to end once the new runtime is deployed to the test chain.

Follow-ons (non-blocking)

• The pool layer (pool.py / pool_client.py) forwards query_mineable_topologies but does not thread an at block parameter (matches the existing query_current_difficulty pattern); admin extrinsics are not surfaced on PoolClient (they use the direct/sudo client).