[MINOR] Default Matrix Multiplication specializations

This commit adds specializations for matrix multiplication with:

1. dense-sparse with sparse output
2. ultra sparse out dense dense in.
3. sparse out on sparse vector right side in.

Furthermore, i modified the call stack to branch to native mm
inside LibMatrixMult, to allow easy native support for CLA, by calling
LibMatrixMult, instead of having to go though a MatrixBlock.

Author: Baunsgaard

src/main/java/org/apache/sysds/runtime/matrix/data/LibMatrixMult.java

@@ -159,6 +159,13 @@ public static MatrixBlock matrixMult(MatrixBlock m1, MatrixBlock m2, MatrixBlock
 	 * @return ret Matrix Block
 	 */
 	public static MatrixBlock matrixMult(MatrixBlock m1, MatrixBlock m2, MatrixBlock ret, int k) {
+		if(NativeHelper.isNativeLibraryLoaded())
+			return LibMatrixNative.matrixMult(m1, m2, ret, k);
+		else
+			return matrixMult(m1, m2, ret, false, k);
+	}
+
+	public static MatrixBlock matrixMultNonNative(MatrixBlock m1, MatrixBlock m2, MatrixBlock ret, int k) {
 		return matrixMult(m1, m2, ret, false, k);
 	}
 

src/main/java/org/apache/sysds/runtime/matrix/data/LibMatrixNative.java

@@ -140,7 +140,7 @@ public static MatrixBlock matrixMult(MatrixBlock m1, MatrixBlock m2, MatrixBlock
 		else
 			LOG.warn("Was valid for native MM but native lib was not loaded");
 
-		return LibMatrixMult.matrixMult(m1, m2, ret, k);
+		return LibMatrixMult.matrixMultNonNative(m1, m2, ret, k);
 	}
 
 	public static void tsmm(MatrixBlock m1, MatrixBlock ret, boolean leftTrans, int k) {

src/main/java/org/apache/sysds/runtime/matrix/data/MatrixBlock.java

@@ -4994,10 +4994,7 @@ public final MatrixBlock aggregateBinaryOperations(MatrixBlock m1, MatrixBlock m
 	public MatrixBlock aggregateBinaryOperations(MatrixBlock m1, MatrixBlock m2, MatrixBlock ret, AggregateBinaryOperator op) {
 		checkAggregateBinaryOperations(m1, m2, op);
 		final int k = op.getNumThreads();
-		if(NativeHelper.isNativeLibraryLoaded())
-			return LibMatrixNative.matrixMult(m1, m2, ret, k);
-		else 
-			return LibMatrixMult.matrixMult(m1, m2, ret, k);
+		return LibMatrixMult.matrixMult(m1, m2, ret, k);
 	}
 
 	protected void checkAggregateBinaryOperations(MatrixBlock m1, MatrixBlock m2, AggregateBinaryOperator op) {