Remove "highly" from "highly optimized C++"

PiperOrigin-RevId: 718940968
Change-Id: I041bd384cd66bbf550bfa4b8b22e2d71528f2adf

Author: dilumich

docs/glossary.md

@@ -367,9 +367,7 @@ The **size** of a DataSlice is the number of items including
 #### Sparsity {#sparsity}
 
 DataSlice can be **sparse**, which means some items in the DataSlice are
-**missing**. Sparsity is supported natively[^1].
-
-[^1]: very sparse case has not been highly optimized yet.
+**missing**. Sparsity is supported natively.
 
 #### Stub {#stub}
 

docs/index.md

@@ -41,8 +41,7 @@ production (i.e. C++ environments).
 *   **Zoom in on your data**: utilize **data views, updates, overlays,
     versions**, including when working with **large data sources**, and use only
     parts of the data needed at the moment.
-*   **Performance**: computation is vectorized and performed in highly optimized
-    C++.
+*   **Performance**: computation is vectorized and performed in optimized C++.
 *   **Evaluate in distributed environment or serve in production**: convert
     evaluation logic into **computational graphs**
     (ASTs) that can be introspected, optimized then evaluated in distributed

docs/overview.md

@@ -39,8 +39,7 @@ Also see [Koda Cheatsheet](cheatsheet.md) for quick references.
 *   **Zoom in on your data**: utilize **data views, updates, overlays,
     versions**, including when working with **large data sources**, and use only
     parts of the data needed at the moment.
-*   **Performance**: computation is vectorized and performed in highly optimized
-    C++.
+*   **Performance**: computation is vectorized and performed in optimized C++.
 *   **Evaluate in distributed environment or serve in production**: convert
     evaluation logic into **computational graphs**
     (ASTs) that can be introspected, optimized then evaluated in distributed