docs: add Future — LIN Bus Simulator planning section to ROADMAP

[SoundMatt]

Summary

• Adds a "Future — LIN Bus Simulator" section to ROADMAP.md, planning-only (no implementation).
• Grounds the plan in the crate's actual current state (verified directly, not from a stale audit note): virtual_bus (684 LOC), mock (273 LOC), master (291 LOC — MasterNode already does real schedule-table execution), slave (175 LOC — thin registration wrapper only), ldf (697 LOC — Db::schedule() already emits the same ScheduleEntry type MasterNode consumes), safety (434 LOC E2E). Confirms zero unsafe in src/ and the tokio/async-trait-based async model.
• Motivates why LIN needs this: no OS-native virtual bus (unlike CAN's vcan) and no third-party LIN stack in this ecosystem to test against (unlike DDS's CycloneDDS peer).
• Scopes a phased plan: Phase 1 minimal in-process sim module (multi-slave, schedule-driven, built on existing VirtualBus/MasterNode/LDF plumbing), Phase 2 fault injection (wrong checksum, silence, wrong PID, delayed response), Phase 3 multi-slave scheduling conflicts + LIN 2.x event-triggered/sporadic slot modeling, Phase 4 (stretch goal) cross-process transport over a Unix domain socket.
• Explicitly distinguishes the simulator's core dev/test value (Phases 1–3, delivered without any interop work) from its potential role as the mechanism for real go-LIN/cpp-LIN interop testing (Phase 4 only), mirroring rust-DDS's two-process rtps-interop-peer pattern.

Test plan

• Docs-only change (ROADMAP.md). No code, no behavior change.
• CI run to confirm the existing pipeline (build/clippy/fmt/test/DCO/RELAY conformance) passes unaffected.