Make rand functions part of JIT

The random functions in are currently not part of JIT. This means that ArrayFire allocates and creates random values for each call of the function before it can be used by other JIT nodes.

With the help of the multi-eval functionality recently added to JIT, we could calculate the random values as part of the JIT engine and write the random values to memory thereby reducing the kernel launch overhead.

先定位 issue 中提到的隨機函式入口點和 JIT multi-eval 實作。追蹤隨機值在進入其他 JIT 節點之前目前是如何配置的；完成的標準是透過 JIT 產生隨機值，同時降低相關的核心啟動額外負擔。

由索引模型根據 Issue 內容生成。