Add dr-inference-engine and route every model session through it

One crate names the runtime, the providers and the devices; dr-face and
dr-segment ask it for a session by role. It hands ort an API table once
per process — from a libonnxruntime it dlopens when the app names a
directory holding one, otherwise from tract — so the Rust build stays
free of C on every target and a package can install the runtime as a
file (docs/inference.md §3).

Sessions live in a registry behind a Model handle that holds the bytes,
not the session: every use refreshes a timestamp and a reaper unloads
whatever sat idle past the decay. A scan that runs the detector on each
image never lets it go idle; a click in the develop view lets the
segmenter go after thirty seconds; a handle used after that reloads,
and reloads on a higher rung if a compiled engine has landed meanwhile.

The probe walks the platform's ladder by building strict sessions and
timing them against the CPU provider, caches the choice against a
fingerprint of the runtime, driver, hardware and models, and compiles
engines for the selected rung in the background, smallest model first.
Nothing in this commit turns the native path on: the apps still run on
tract until they call init with a runtime directory.

core/dr-inference-engine/Cargo.toml

@@ -0,0 +1,44 @@
+[package]
+name = "dr-inference-engine"
+version.workspace = true
+edition.workspace = true
+rust-version.workspace = true
+license.workspace = true
+
+# The one crate that names a runtime, a provider, a vendor library or a
+# device (docs/inference.md §8). `dr-face` and `dr-segment` ask it for a
+# session by role and never see which of these answered.
+
+[dependencies]
+thiserror.workspace = true
+log.workspace = true
+serde.workspace = true
+serde_json.workspace = true
+
+# `ort` is the API; what supplies it is decided once per process (§3):
+# `libonnxruntime` found on disk, or `tract`. Both are behind
+# `alternative-backend`, so nothing here links C on any target.
+ort = { workspace = true }
+ort-tract = { workspace = true, optional = true }
+# dlopen, and the C types of the table it fetches. Both pure Rust;
+# `libloading` is already in the tree through wgpu.
+libloading = { version = "0.8", optional = true }
+ort-sys = { version = "2.0.0-rc.13", default-features = false, features = ["disable-linking"], optional = true }
+
+# The NVIDIA rungs exist on the desktop only. These features add `ort`'s
+# option builders and nothing else — no linking under `alternative-backend` —
+# but an Android binary has no business carrying even the option names, and
+# the packaging must never be tempted to (§2, §3.1).
+[target.'cfg(not(target_os = "android"))'.dependencies]
+ort = { workspace = true, features = ["cuda", "tensorrt"] }
+
+[target.'cfg(target_os = "android")'.dependencies]
+ort = { workspace = true, features = ["qnn"] }
+
+[features]
+# The floor: `tract` supplies the API table when no runtime file is found, or
+# always, in a build without `native`. Tests want this and nothing else.
+default = ["tract"]
+tract = ["dep:ort-tract"]
+# Look for `libonnxruntime` on disk and hand its table to `ort`.
+native = ["dep:libloading", "dep:ort-sys"]