Implement memory management for the standalone functions cache

[dumbbell]

In #72, we introduce a cache for already extracted standalone functions to improve performances.

Currently, this cache will grow indefinitely. We need some kind of policy to clean it up.

[the-mikedavis]

What would you think about an ETS-based LRU cache for this? Would you be open to a contribution?

[michaelklishin]

In other RabbitMQ projects where we have a cache, it is sometimes an interface that you can implement however you like and configure a module to be used. Maybe it's an overkill for Khepri but there is no single best way to do K/V caching in process, so an interface makes certain sense to me. Our first implementation can be ETS-based :)

[dumbbell]

What would you think about an ETS-based LRU cache for this? Would you be open to a contribution?

Yes, feel free to experiment!