doc(readme): mark M9 (#139) complete + refresh roadmap & nft_marker_gen

[kalwalt]

Summary

Documentation + one small Cargo.toml follow-up to PR #176 (M9 merge). Marks the M9 milestone as complete in the README roadmap, refreshes the post-M9 short-term goals, brings the Project Structure section in line with the new pluggable backend architecture, and corrects the nft_marker_gen example invocation now that the pure-Rust default actually works for it.

Closes the lingering documentation TODO from #139.

Changes

README.md

• Roadmap → Completed Milestones: M9 (feat(kpm): M9 - VisualDatabase and pure Rust backend #139) added with full sub-milestone breakdown:
  • M9-1 / feat(kpm): M9-1 - port visual_database.h to freak/visual_database.rs #140 — VisualDatabase port
  • M9-2 / feat(kpm): M9-2 - implement RustFreakMatcher and DualFreakMatcher #141 — pluggable FreakMatcherBackend trait + RustFreakMatcher / DualFreakMatcher
  • M9-3 / feat(kpm): M9-3 - remove FFI as default, pure Rust backend complete #142 — pure-Rust as default (no C++ toolchain required)
  • Plus cross-cutting wins: cross-platform/cross-stack matcher determinism, hand-annotated absolute corner-error gate, cross-stack parity gate vs jsartoolkitNFT@1.10.0.
• Roadmap → Short-term Goals (toward v1.0.0): replaced the now-stale "Complete KPM in idiomatic Rust" bullet with the actual pending work:
  • KPM-specific Criterion benchmark (deferred from feat(kpm): M9-3 - remove FFI as default, pure Rust backend complete #142)
  • WASM browser examples (feat(wasm): full WASM support + browser examples after M9-3 #161)
  • M9 patch coverage uplift 84.76% → ≥90% (test(kpm): raise patch coverage on M9 modules (84.76% → target ≥90%) #177)
  • Criterion 0.5 → 0.8 (chore(deps): upgrade criterion 0.5.1 → 0.8.x #174)
• Project Structure: kpm module section now describes the pluggable FreakMatcherBackend trait, the pure-Rust default since M9-3, the new kpm::rust_backend / kpm::cpp_backend / kpm::freak::visual_database modules, and the BTreeMap-for-determinism note on kpm::freak::hough.
• NFT Marker Generation Example: clarified that the example builds on the pure-Rust default and produces .iset + .fset; ffi-backend is opt-in for the .fset3 (FREAK descriptors) step. Two invocations shown, plus a what-you-get table.

crates/core/Cargo.toml

• Relax nft_marker_gen required-features from ["log-helpers", "ffi-backend"] to ["log-helpers"].
• The source already had #[cfg(not(feature = "ffi-backend"))] branches that skip the .fset3 step with a warning — Cargo.toml was the only thing blocking running it on the pure-Rust default. The C++ .fset3 path still works exactly as before when ffi-backend is enabled.

What this PR does NOT do

• It does not wire nft_marker_gen to use RustFreakMatcher for .fset3 generation. The .fset3 step still calls CppFreakMatcher when ffi-backend is enabled. Doing that properly needs its own PR with byte-level parity validation against existing .fset3 files — filed as the planned follow-up referenced in the new docs.
• No source-code changes to anything other than the example's feature gating.

Test plan

• cargo check -p webarkitlib-rs --example nft_marker_gen --features log-helpers (pure-Rust path) succeeds.
• cargo check -p webarkitlib-rs --example nft_marker_gen --features "log-helpers,ffi-backend" (legacy path) succeeds.
• All existing CI jobs stay green.
• README renders correctly on GitHub (anchors, tables, code fences).

Risk

Low. Pure documentation changes plus one feature-gate relaxation. The Cargo change is strictly less restrictive — anyone previously running with ffi-backend keeps the same behavior; users who didn't have a C++ toolchain can now run the minimal .iset + .fset path that the source code already supported.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!