vtracer-bench: codec-free library, codecs behind a cli feature

- The library now speaks the image crate's buffer types only:
  fidelity(&RgbImage, &RgbImage, thresh) -> (FidelityReport, GrayImage).
  A tracer embedding the metric pulls no image codec at all.
- New `cli` feature (default on) brings png/jpeg/webp decoding and
  gates the vtracer-bench bin (required-features).
- dssim-core with default features off: drops the rayon thread pool.
- Scores are unchanged: the CLI prints the same value before and after.

crates/vtracer-bench/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "vtracer-bench"
 description = "Blind fidelity benchmark for raster-to-vector tracers: compare the original raster with a rendered reconstruction and get one 0..1 fidelity score built from PSNR, SSIM and a clustered-diff patch metric."
-version.workspace = true
+version = "1.0.0-alpha.5"
 authors.workspace = true
 edition.workspace = true
 license.workspace = true
@@ -10,9 +10,19 @@ repository.workspace = true
 categories = ["graphics", "development-tools::testing"]
 keywords = ["vectorization", "benchmark", "fidelity", "ssim", "psnr"]
 
+[features]
+default = ["cli"]
+# The CLI decodes originals and rendered candidates from files; the
+# library itself only speaks the image crate's buffer types, so a
+# tracer embedding the metric pulls no codec at all.
+cli = ["image/png", "image/jpeg", "image/webp"]
+
+[[bin]]
+name = "vtracer-bench"
+required-features = ["cli"]
+
 [dependencies]
 visioncortex.workspace = true
-dssim-core = "3"
+dssim-core = { version = "3", default-features = false }
 rgb = "0.8"
-# Decode-only: trimmed to real input formats (drops the AV1 encoder + OpenEXR).
-image = { version = "0.25", default-features = false, features = ["png", "jpeg", "webp"] }
+image = { version = "0.25", default-features = false }