Add mkl support for matrix multiply. (#86)

* Fix some rebase issues.

* Use mkl instead.

* Use mkl in bert.

* Add the optional mkl feature.

* Conditional compilation based on the mkl feature.

* Add more mkl support.

candle-examples/examples/bert/main.rs

@@ -1,4 +1,8 @@
 #![allow(dead_code)]
+
+#[cfg(feature = "mkl")]
+extern crate intel_mkl_src;
+
 use anyhow::{anyhow, Error as E, Result};
 use candle::{safetensors::SafeTensors, DType, Device, Shape, Tensor};
 use candle_hub::{api::Api, Cache, Repo, RepoType};