[FEA] Improvements to benchmarking methodology in the docs

The following feedback was given about our benchmarking methodology page in the docs. We should strive to improve these docs so they are more clear and easier to understand.

1. Pls define exact vs. actual nearest neighbors. Is exact the set known by an oracle, e.g. with exhaustive search, and actual is what the heuristic search yielded?

2. The (intersection of 2 sets of K)/K method evaluates wrt the exact set, but other recall evaluation methods consider whether you're within a given convergence threshold, correct? What's that called, and is that relevant enough to bring up here?

3. Meaning of grey line isn't clear, or why green lines don't go all the way to 1

4. Bucket list shows 85% but graph shows 80%