Graph tests

[Zannick]

After adding many new details to a graph, I've been running the whole program to see if it succeeds at getting to the latest items, but when it doesn't, it takes quite a deal of work to find out why--I've used the visual graph a few times to find places where I've missed edges.

Ideally we would have a better solution for this. In keeping with the codegen, we could write our tests in yaml, at least the ones that are formulaic:

- with: [items...]
  context:
    foo: bar
  requires: item
  goal: Spot or item

Some ideas for tests:

• any_path_to to check that the given setup can reach the target spot
• no_path_to for the opposite
• can_obtain to test collecting an item (at any location available)
• requires to test both any_path_to and no_path_to with/without a specific item
• path to check a specific path

The generated test code should be able to pinpoint where in the graph the expectation fails (which would be different based on the type of test, for example), or at least be more descriptive than simply "panicked because it couldn't reach the objective".

[Zannick]

Structure of this is going to be macros and functions defined in analyzer and codegen a test file with all the tests.

It'd be hard to find specific exits that don't exist, except in the case of a specific path. So for the proposed any_path_to, we'd output all steps from a spot we can access to a spot we can't. (Unfortunately this includes testing warps at every spot.)

Currently we can't simulate having other movement types, but that could be done with a refactoring of local_travel_time.