The Witness: Change all option name comparisons to strings instead of numeric values

[NewSoupVi]

Refactor all the options checking to do world.options.shuffle_EPs == "individual" instead of world.options.shuffle_EPs == 1

Also, do a refactor of the shuffle_postgame code, which is where most of the option checking happens. This part probably needs to be looked over by someone who understands the game. I'm gonna request a review from Exempt-Medic I think.

[NewSoupVi]

This will likely conflict with #2499 so I'll wait for that before un-drafting this one, bc that one is more important

[Exempt-Medic]

I used a regex search and found some other comparisons that could be strings and some simplify-able comparisons:

• hints
• player_logic
• player_logic
• player_logic
• player_logic
• __init__
• __init__
• items
• items
• items
• static_logic
• utils

[NewSoupVi]

I'm happy with shuffle_doors >= 2

The line
+= shuffle_doors > 1
is necessary because the result needs to be exactly 1 or 0 numerically.

Some of the code is not mine and I'm worried that removing "> 0" could cause some problem with negative values and it's outside of the scope of this PR anyway

As for the two remaining things relating to options, I've changed those now :) Thanks

[alwaysintreble]

Some of the code is not mine and I'm worried that removing "> 0" could cause some problem with negative values and it's outside of the scope of this PR anyway

just peeked at it and .value would be better/faster if it's a toggle. if it's a choice it's fine

[Exempt-Medic]

I read through the changes, primarily to the postgame section, to check that they lined up with my understanding of the game. I then ran 1500 fully-random, differently-seeded generations on main and option_names and compared the outputs. I also specifically checked all 96 reasonable combinations of postgame settings. The outputs were identical except when the one additional postgame consideration occurred (lines 324-235 of player_logic.py), which is the expected behavior.

[NewSoupVi]

This PR is at the bottom of a pretty big dependency chain at this point, so it'd be epic if this could be prioritised over other Witness PRs 🙏 It shouldn't be super hard to review I think

[black-sliver]

LGTM

Having any kind of unit test that verifies the options put the corresponding items/things into the pool/member variables, and that slot data has the correct output of the correct input wouldn't hurt going forward though. (It would make sure it behaves the same before and after, and save you from typos in strings)