web(#919): per-node local_hashps series in /local_stats, honest-absent when cold

[frstrtr]

#919 — per-node local hashrate series in /local_stats (honest-absent)

Problem

/api/local_stats exposed no per-node local hashrate. The only hashrate field was poolhashps (pool-wide attempts/s — the "Pool: X GH/s" figure). A local-hashrate graph therefore had no real series of this node's own work rate to poll.

Fix (additive)

Add local_hashps to MiningInterface::rest_local_stats (src/core/web_server.cpp), sourced from the real stratum work-rate counter m_stratum_hashrate_fn — the same live per-node source the node-fee local_hr leg already trusts, not the empty m_node stub that produced the founding poolhashps=0 lie.

Honest-absent, not flat zero: with no work-rate source wired the counter is cold and the field is null — an honest "unknown", never a 0 that would falsely claim the node mines nothing. A wired source reporting 0.0 is a true reading (no active local miners) and stays 0.

Every existing field is byte-identical; only the new key is added.

Source note for the merge gate

local_hashps is this node's local work rate (stratum counter), deliberately distinct from poolhashps (pool-wide). If the intent was instead to mirror the pool counter per-node, say so and I will repoint the source — the seam is one line.

Tests

src/core/test/web_server_local_hashps_test.cpp — 2 additive KATs, folded into the existing core_test target:

• ColdCounterIsHonestNullNotZero — no source wired → local_hashps present and null (not 0).
• LiveCounterSurfacesRealValue — source wired → local_hashps equals the real counter value.

Both FAIL without the source change (key absent → contains("local_hashps") is false), verified by reverting only web_server.cpp and rebuilding. With the fix: core_test 145/145 green.