feat(#593): opt-in local miner — hand off stratum URL + secret for a co-located RigForge worker

[VijitSingh97]

Closes #593.

What this does

Adds an opt-in local miner: a box that runs the stack 24/7 can mine with its spare CPU by co-locating a RigForge worker on the stack host, pointed at the stack's own loopback stratum. The dashboard already discovers workers from the proxy, so a local rig self-registers with no stack-side wiring — the only friction was copying two values by hand. This closes that gap.

Scope (deliberately narrowed to config + hand-off + docs)

Per the issue's re-scoped design, RigForge owns all host-level tuning (HugePages, GRUB, MSR, governor, the miner service). Pithead's job is purely to declare intent and hand off. This PR does not vendor/drive RigForge or render a RigForge config.json — those remain follow-on work; a manual RigForge install on the same host, fed the two printed values, is fully functional today.

Pithead-side wiring added:

• local_miner.enabled config flag (default off), registered in config.reference.json to satisfy the closed-schema superset invariant (feat(dashboard): commit dashboard.energy from the config editor + close config.json smuggling gap in the control gate (#504, #33) #537/tier 1 drift guard: every config path pithead reads must exist in config.reference.json #561). The drift-guard test picks up the new config_bool read site automatically (still green).
• pithead setup prompt — "Also mine on this machine with its spare CPU?" (default N), wired through wizard_ask_shape → wizard_write_config like the existing clearnet/onion/telegram opt-ins.
• announce_local_miner() — on setup and apply, surfaces the two values a RigForge install would otherwise prompt for:
  • Pool URL: 127.0.0.1:${STRATUM_PORT} for a 0.0.0.0/loopback bind, or the configured p2pool.stratum_bind address for the specific-LAN-bind edge case the issue calls out (loopback wouldn't reach it).
  • Stratum secret: PROXY_STRATUM_PASSWORD from .env, shown only when set (a "none set" note otherwise).
• Docs (docs/workers.md#mine-on-the-stack-host-itself) — how to enable it, that RigForge owns the host tuning, and the host-global-governor / thread-cap caveats.

No HugePage / GRUB / nr_hugepages / MSR / governor code was added — confirmed. (Note: the issue's claim that Pithead has no such code is inaccurate — optimize_kernel already does GRUB/HugePages tuning for the stack's own RandomX use — but this PR adds none and does not touch it.)

Testing

Tier-1 shell (tests/stack/run.sh), all green (1550 passed):

• announce_local_miner hands off the stratum URL + secret only when opted in (#593) — opt-out prints nothing, absent block prints nothing, opt-in surfaces the loopback pool URL + secret + the RigForge-owns-tuning note, LAN-bind targets the bound address:port, and the no-password case is stated explicitly.
• wizard_ask_shape has exactly 6 read prompts — prompt-count pin updated (5 → 6).
• defaults path: no local_miner block written (opt-in off, #593) and opt-in path: local_miner.enabled written true (#593) — wizard prompt → write wiring.
• config.reference.json stays a complete superset (tier 1 drift guard: every config path pithead reads must exist in config.reference.json #561) — passes with the new key.

Tier-4 deferred: a real co-located mining run (start a thread-capped RigForge worker against the stack's stratum, verify it self-registers and the stack keeps sync/share reliability) needs a mining host and is out of band for this config/doc change — matches the issue's own validation plan.

make lint: all touched surfaces pass (shellcheck/shfmt, docs-voice 31/31, markdownlint 0 errors). lint-proto fails locally only because it requires a running Docker daemon (Tari protobuf lint, unrelated to this change).

🤖 Generated with Claude Code