Enable cloning in memory to speed up rive load when using runtime asset swapping

[tguerin]

When using AssetLoader for runtime asset swapping, there is no straightforward way to reuse a RiveFile. Each time it's needed, it must be re-imported. While debugging performance, I noticed a significant amount of time was spent in the _readRuntimeObject method, leading to inefficiencies.

To address this, I implemented a caching mechanism to store RiveFile objects and enable cloning them directly in memory. This eliminates the need for repetitive re-imports and significantly reduces loading overhead.

This change resulted in a substantial improvement in loading times, with a performance boost of 4x to 10x depending on the use case.

What do you think about this approach? Are there any additional considerations or potential pitfalls I should address?

[HayesGordon]

This is great, @tguerin; we'll get this reviewed as soon as possible.