Build the standalone native resting-order state machine

[luisleo526]

Native strategies previously had queued market actions but no persistent resting-order lifecycle. This PR adds that state machine to the standalone C++ host: limit, stop, stop-limit and trailing requests; partial execution with capacity per input point; owner-bound children; and causal group cancellation/reduction through general submit, replace and cancel commands.

Confirmed OHLC and forward ticks use the same working-request core. Matching chooses the earliest remaining price crossing, preserves exact reached prices across rescans, and keeps observed gaps distinct from modeled continuous paths. Committed opening events establish child ownership and quantity. Cycle changes, replacement, exhaustion, group effects and dependency cleanup have deterministic terminal behavior.

The existing physical lot book remains the accounting authority. Scoped closes preserve unrelated lots and historical costs with one current fee per execution. Typed immutable events retain partial quantities, activation, ownership, receipts and failure prefixes; failed runs require discard/replay.

Requests/core/events use native_order_v2; identity values remain v1. Engine, pending, host and consumer C++ boundaries move together to epoch13, with broker/stream hash domains and old return-only native_events() link controls. C ABI4 prefixes and current calendar/session/timezone/DST/timeframe support are preserved. Complete Pine policy lowering remains codegen/adapter work in R4.

Validation on exact final candidate c57e8687d91027df09474b414369162e47b94bbb, tree e209c89035ec884774f8f3e68e1dc118f26aaf51:

• Full local suite: 326 tests passed, zero failed; one WebSocket test skipped because the selected macOS libcurl lacks WebSocket support. All nine final PR CI checks passed, including Linux/macOS, sanitizers, and the native live runner.
• Independent native witness ledger maps every v5 V/C/T/O/G/S/P requirement to host, core, or type evidence. Runtime, ABI/hash, test-oracle, and final scope reviews are GREEN. Concrete regression witnesses cover every discovered defect.
• Cloud experiment exp-native-r2-c57e868-20260912: 72/72 cases, 4,190/4,190 probes, both provider executions successful. Zero full-grade, raw-trade, substantive-verifier, configuration, metadata, or coverage differences against the accepted R2a reference. Raw verifier metadata differences are retained under the existing comparator rules.
• Actual gate pineforge-pr-gate-n976j: FAIL, target.not-positive only. Unchanged compatibility is accepted under the standing neutral-refactor authorization; no PASS receipt or baseline promotion is claimed.
• The sanitizer ABI test now supplies historical v12 RTTI without disabling checks. The existing crash-recovery test uses a fresh writer process, preserves abrupt-exit recovery assertions, and passed 50 consecutive runs.

Squash merge must preserve the exact reviewed/measured tree. R3/R3b accounting consolidation, R4 source-policy ownership, final audit, and the later parity campaign remain subsequent roadmap work.