[C++ backend] RoboPlan adapter (native solveInto)

[siddhss5]

Sub-issue of the native C++ backend umbrella #482.

Wire the native allocation-free solveInto (#485) into RoboPlan's in-loop IK — the first real consumer of a native artifact.

• Thin adapter: RoboPlan endpoint pose -> Pose, caller-owned std::span<Solution<N>>, SolveOptions from the planner's limits/seed.
• Validates the allocation-free path under repeated in-loop calls.
• Full payoff arrives once RoboPlan also has native endpoint-region sampling (tracked separately on the RoboPlan side).

Depends on Phase 0a (#488) + ABI (#485).

[siddhss5]

Deferred / de-emphasized. Do not start until the native artifact is bulletproof — Phases 0a-3 (#488/#489/#490/#491) conformance (#487) green across the shipped solver families. Wiring a planner onto an unproven artifact obscures where a failure lives. This is no longer 'Phase 0b'; it starts only after the bulletproof-C++ gate.

[siddhss5]

Deferral lifted — the "bulletproof C++" precondition is met (72/72 native + parity-gated #556/#558; crash-hardened + native-default #560, v5.0.0; every arm emits a self-contained cpp/gen/<arm>.hpp). Builds on the bare-C++ consumer path/docs (#561). Part of the #554 effortless-C++ pillar (D), alongside #493 (MoveIt plugin) and #561 (consumer docs).

[siddhss5]

Distribution model: upstream a thin adapter into RoboPlan IF it exposes an IK-backend hook and will take the dependency; otherwise host the adapter in ssik.

The <arm>.hpp is the stable ABI, so the adapter is thin either way. Precedent for upstreaming into the consumer: the Alex V2 / IHMC work (deliverable was a PR into the consumer's SDK, not an ssik prebuilt).

Deciding question (blocks scoping): does RoboPlan expose a pluggable IK-backend interface (a solveInto-shaped hook), and is it willing to depend on / vendor the ssik header?

• Yes → PR a thin solveInto adapter into RoboPlan; they version it with their releases.
• No → host the adapter in ssik against RoboPlan's public API.

Builds on the bare-C++ consumer path/docs (#561).