Keep the embedder off the Hexagon, and let the probe example ask for a runtime

On the tablet the engine compiled arcface for the NPU: the routing
compared the form a rung wants with the form on offer, and for the
embedder both are f32, so nothing said no. A rung now says which roles
it serves at all, and the Hexagon does not serve the embedder (§7 —
its vectors must compare across devices). Tested at the routing seam.

dr-segment's onnx_probe example still named ort-tract, which is what
stopped the workspace test build.

core/dr-inference-engine/src/engines.rs

@@ -55,6 +55,7 @@ pub fn run() {
     let mut jobs: Vec<(crate::Role, Source, u64)> = cfg
         .models
         .iter()
+        .filter(|(role, _)| rung.serves(*role))
         .filter_map(|(role, path)| {
             let (path, form) = crate::resolve_model(*role, path);
             (form == rung.form(*role)).then(|| {
@@ -65,7 +66,7 @@ pub fn run() {
         .chain(cfg.embedded.iter().filter_map(|(role, bytes)| {
             // An embedded model has no int8 sibling to offer a rung that
             // wants one; it runs on that rung's fallback.
-            (rung.form(*role) == Form::F32).then_some((
+            (rung.serves(*role) && rung.form(*role) == Form::F32).then_some((
                 *role,
                 Source::Bytes(bytes),
                 bytes.len() as u64,