Group inline snapshot transform results by env

[billyjanitsch]

Slightly different, less polished approach to #117.

Not sure if this is worth the additional layer of abstraction, but it solves a few problems in that PR:

• Every test by default has to remember to add an assertion for each env. It would be easy to miss one, especially if the snapshots are too large to all fit on one screen. This approach defaults to transpiling for every env.
• There's no easy way to tell whether two envs produce the same transpilation result without a close comparison. It's easy to miss subtle differences. This solves that by grouping envs that product the same result. It could also be solved by doing expect(env1Output).toBe(env2Output) but we don't necessarily want to declare as an assertion that these should always be the same, we just want to know if they are and when that changes.