documenting benchmarks

[AboudyKreidieh]

In order for benchmarks to be clear for people, we require tutorials on the following:

• importing and running benchmarks
• evaluating and publishing results

In addition, we need to document each of our benchmarks on the website, maybe in the style of OpenAI gym (see https://gym.openai.com/envs/#classic_control):

• bottleneck
• grid
• figure eight
• merge

[eugenevinitsky]

All of this stuff actually winds up under the purview of Flow Garden; we're going to need all of this for its first release. We should do a first pass of this before the paper is out, but then we'll shift its location