Rules unit tests: Do static analysis instead of fuzz testing

[MatthewMarinets]

What task needs to be completed?

The new test that checks all items used in rules are marked progression is a fuzz test, which unfortunately means it's very slow and probably shouldn't be run with the other unit tests.

Following a discussion in the discord (starting around here), we evaluated some options and suggested that the best course of action to get an automated test with reasonable sanity-checking and performance would be a test that looks at the contents of the file. At a basic level, the changes should:

• Move defense-rating dicts from items.py to rules.py
• Make a test that scans the contents of rules.py, searching for instances of item_names.(\w+) and checking that the associated item is progression
  • To be really fancy, this parser probably needs to account for comments, which also means accounting for strings. May not be necessary given item_names.(\w+) probably won't appear too much in strings and comments, and we can just check that anyways
• Make tests that also scan the contents of the starting_unit / basic_unit lists in items.py

Other possibilities would include

• some kind of intelligent depth-first search algorithm to explore every path of every rule (a lot of work, very complicated for a test)
• Reduce the repetitions on the fuzz test and just rely on it running frequently (would potentially lead to flaky tests, though)
• Something slower / manually initiated, run before releases / milestones?