feat: add embedding model continuous batching scheduler (#564)

* feat: add embedding benchmark and implement continuous batching

Signed-off-by: Huamin Chen <hchen@redhat.com>

* add embedding examples

Signed-off-by: Huamin Chen <hchen@redhat.com>

* expose the batch api

Signed-off-by: Huamin Chen <hchen@redhat.com>

* review feedback

Signed-off-by: Huamin Chen <hchen@redhat.com>

* review feedback

Signed-off-by: Huamin Chen <hchen@redhat.com>

* review feedback

Signed-off-by: Huamin Chen <hchen@redhat.com>

* fix precommit

Signed-off-by: Huamin Chen <hchen@redhat.com>

* review feedback

Signed-off-by: Huamin Chen <hchen@redhat.com>

---------

Signed-off-by: Huamin Chen <hchen@redhat.com>
Signed-off-by: Huamin Chen <rootfs@users.noreply.github.com>

Author: rootfs

candle-binding/Cargo.lock

@@ -7,7 +7,8 @@ license = "MIT OR Apache-2.0"
 
 [lib]
 name = "candle_semantic_router"
-crate-type = ["staticlib", "cdylib", "rlib"]
+# Order: rlib for Rust-to-Rust linking, staticlib for C/Go FFI (static), cdylib for dynamic linking
+crate-type = ["rlib", "staticlib", "cdylib"]
 
 [features]
 default = ["cuda"]
@@ -38,6 +39,7 @@ rand = "0.8.5"
 rayon = "1.8"       
 once_cell = "1.19"  # Used for lazy initialization of global state
 parking_lot = "0.12"
+crossbeam-channel = "0.5"  # Efficient multi-channel select for scheduler wakeup
 
 [dev-dependencies]
 rstest = "0.18"