Fp64 on the CPU (#1843)

* add fp64 data type

* clean build

* update docs

* fix bug

mlx/backend/cpu/gemms/cblas.cpp

@@ -41,4 +41,39 @@ void matmul<float>(
   }
 }
 
+template <>
+void matmul<double>(
+    const array& a,
+    const array& b,
+    array& out,
+    bool a_transposed,
+    bool b_transposed,
+    size_t lda,
+    size_t ldb,
+    float alpha,
+    float beta) {
+  size_t M = a.shape(-2);
+  size_t N = b.shape(-1);
+  size_t K = a.shape(-1);
+
+  for (int i = 0; i < (a.size() / (M * K)); ++i) {
+    cblas_dgemm(
+        CblasRowMajor,
+        a_transposed ? CblasTrans : CblasNoTrans, // transA
+        b_transposed ? CblasTrans : CblasNoTrans, // transB
+        M,
+        N,
+        K,
+        alpha, // alpha
+        a.data<double>() + elem_to_loc(M * K * i, a.shape(), a.strides()),
+        lda,
+        b.data<double>() + elem_to_loc(K * N * i, b.shape(), b.strides()),
+        ldb,
+        beta, // beta
+        out.data<double>() + M * N * i,
+        out.shape(-1) // ldc
+    );
+  }
+}
+
 } // namespace mlx::core