feat(scanner): Resolve Askama template dependencies

[reneleonhardt]

What does this PR do?

Record file dependencies introduced by literal Askama #[template(path = "...")] attributes. Paths reuse the existing Rust literal decoder and Cargo package ownership, resolve under the package's default templates/ directory, and are accepted only for one exact configured, indexed in-repository target. Dynamic, malformed, missing, external, excluded, escaping, ambiguous, and unowned routes remain unresolved. Packages with askama.toml and attributes with config = ... also fail closed; custom template-directory configuration is intentionally partial coverage.

Type of change

• Bug fix
• New feature
• New language support
• Documentation
• Other (describe below)

Checklist

• I've tested this locally with go build && ./codemap .
• I've read CONTRIBUTING.md; this does not add a new language.
• Documentation is unchanged because the dependency output is self-describing.

Additional notes

Focused extraction and conservative-resolution tests, the scanner suite, and staticcheck are GREEN.

Real-world Worktrunk smoke tests recover all six template dependencies from src/shell/mod.rs to the Bash, Fish, Fish wrapper, Nushell, PowerShell, and Zsh files under templates/.

This branch targets and contains no generic proc-macro, Cargo-topology, configuration, or non-Rust changes.

Developed with carefully directed, manually reviewed AI assistance.

[JordanCoin]

Verified on real Cargo crates. The property that makes this safe: resolution is existence-gated — an edge is only emitted when a file sits at exactly <pkg>/templates/<literal> and is uniquely indexed, with files[0] != target guarding against the index's extension-stripped keys. Tested the case that matters most: a crate with askama.toml pointing at views/ plus a decoy at templates/hello.html produces no edge at all rather than guessing the decoy. #[template(source = "...")] with a same-named decoy file also correctly produces nothing.

On scope — it's a fair question whether codemap should carry per-crate support, and I think this earns it: #[template(path=...)] against <crate>/templates/ is shared by askama, rinja, sailfish and yarte, so it's closer to a Rust convention than one library. And a special case that can only fail closed is much cheaper to own than one that guesses.

Two follow-ups, neither blocking: path must be the first attribute argument, so #[template(escape = "none", path = "x.html")] is silently missed (one more pattern fixes it); and on Windows filepath.IsAbs("/foo.html") is false, so a leading-slash literal resolves there while it's correctly rejected on Unix — a cross-platform divergence worth closing with a strings.HasPrefix(value, "/") check.

If a second template crate ever arrives, please table-drive it (attribute name, default dir, config filename) rather than adding a second resolver — the hard-coded askama.toml check is already the seam.