Cap concurrent Electrum checkpoint Merkle proof builds

[agoodminute]

Summary

get_header_merkle_proof builds a checkpoint Merkle proof — walking and hashing every header from genesis up to cp_height — whenever an Electrum request sets a non-zero cp_height. Until now there was no limit on how many of these builds could run at once; each is comparatively expensive, and every concurrent call added its own CPU cost with no coordination between connections.

Why

Since each build's cost scales with cp_height and there was no shared limit, a burst of concurrent requests could add up to significant CPU work across cores. Capping how many run at once keeps that work bounded regardless of how many connections request it simultaneously.

Changes

• get_header_merkle_proof now takes a checkpoint_proof_concurrency_limit parameter and acquires a permit before building the proof, so at most that many builds run at once, process-wide.
• New config option --electrum-checkpoint-proof-concurrency-limit (electrum_checkpoint_proof_concurrency_limit), default 2. A request past the limit fails immediately rather than queuing behind in-flight builds. 0 rejects all such requests.
• Threaded the new limit through Connection::new / RPC startup alongside the existing txs_limit / subscription_limit.

[EddieHouston]

lgtm b9203b6

[EddieHouston]

b391e2f lgtm