feat: scaffold Rust crate (edition 2024, serde + jiff)

[FabioRocha231]

Slice 0 / 11 of the bottom-up Rust rewrite tracked in #27 — the crate scaffold.

Each later slice lands one module + its tests against a green build, in dependency order, so CI stays green and any regression is trivial to bisect. This PR is just the skeleton.

• rust/ single binary crate tokenline, flat src/, edition 2024, MSRV 1.85.
• Deps: serde + serde_json + jiff (jiff::Timestamp for ISO-8601 → epoch, first used in slice 3).
• panic = "unwind" in the release profile — the final slice's top-level catch_unwind (never-crash) depends on it.
• main.rs is a stub, lib.rs is the module root. cargo build / run / test green (0 tests).

Stacked series — this is the base; subsequent slice PRs target their predecessor and retarget to main as each merges.

[ropdias]

Reviewed and tested locally on WSL2 (rustc 1.96.1). All gates green:

• cargo build --locked ✅
• cargo test --locked ✅ (0 tests, as expected)
• cargo clippy --locked --all-targets -- -D warnings ✅ (no warnings)
• cargo fmt --check ✅ (no diffs)
• cargo build --release --locked ✅ → 287 KB stripped binary, runs and exits 0

Clean, well-scoped scaffold. A few things I really liked:

• The bin + lib split from day one — pure, golden-testable domain in lib.rs
  and a thin stateful edge in main.rs. Exactly the shape the rewrite needs.
• panic = "unwind" kept explicitly in the release profile, against the usual
  "add panic=abort to shrink the binary" advice, because the never-crash design
  depends on catch_unwind. The inline comment makes the intent unmistakable.
• edition = "2024" and rust-version = "1.85" are internally consistent, and
  Cargo.lock is committed (correct for a binary).

Solid base for the stack. Approving. 🚀