Add simple caching into enrichers

[sungam3r]

Also fixes readme formatting.

Before:

After:

[sungam3r]

@nblumhardt This caching will manifest itself the more, the less the switching of threads in the application. In a single-threaded application there will be 100% hits. It does not add a lot of load.

[nblumhardt]

Very interesting! The lack of synchronization on _lastValue is a sneaky optimization... 😄 (as you'd have noted, if the value ends up cached in a register and isn't updated from memory despite concurrent writes by other threads, the hit rate seems like it might actually be higher than expected.)

My gut feel is that this will be a positive, it's tricky to gauge the impact without benchmarks, but then, it's also nearly impossible to benchmark meaningfully without a huge amount of effort. Let's do it! 👍

[sungam3r]

it's tricky to gauge the impact without benchmarks, but then, it's also nearly impossible to benchmark meaningfully without a huge amount of effort.

Well, it is possible. But in this respect, and without a benchmark, it is clear that there will be fewer allocations, the whole point is in your own desire to see a performance increase. 5, 10, or 99% reduction in allocations - the specific number is not so important, it can vary from scenario. The approach itself is important - do not allocate the same object over and over again.