Add some preliminary ONNX support (#1260)

* Add the onnx protos.

* Move the reading bits.

* Install protoc on the CI.

* Install protoc on the cuda CI too.

* Use clap for the onnx tool.

* Tweak the CI protoc install.

* Add some simple evalution function.

* Add some binary operator support.

candle-onnx/src/lib.rs

@@ -0,0 +1,14 @@
+use candle::Result;
+use prost::Message;
+
+pub mod onnx {
+    include!(concat!(env!("OUT_DIR"), "/onnx.rs"));
+}
+
+mod eval;
+pub use eval::simple_eval;
+
+pub fn read_file<P: AsRef<std::path::Path>>(p: P) -> Result<onnx::ModelProto> {
+    let buf = std::fs::read(p)?;
+    onnx::ModelProto::decode(buf.as_slice()).map_err(candle::Error::wrap)
+}