Row-based cache invalidation

[mostafa]

The current implementation of the caching stores tables as keys to invalidate the responses. This means that if any change is made to a table (or more), this will invalidate the entire cache for that table. This is inefficient, as read-heavy tables would require unnecessary caching and invalidation (aka. heavy-lifting) between the GatewayD, the cache plugin, the cache server (Redis) and the database (PostgreSQL). This means that for every update to even a single row in a table that has many cached responses in the cache server, the plugin will remove all the cached responses, effectively making subsequent queries to hit the database for a while until all those cache responses are cached.