Fix eject pack interactions

[ShivaD173]

• Fix Commander not taking hazard damage
• Fix some eject pack interactions
  • Applying effects (and switching out) is speed based instead of location based
  • Cannot switch out until pokemon have started
• Still to be fixed:
  • Eject Button should trigger before Eject Pack
  • Log should only display eject pack on switch, not when it is triggered
  • Whatever the last test in ejectpack.js is supposed to test.

[DaWoblefet]

• Whatever the last test in ejectpack.js is supposed to test.

This test is definitely off; what I think I was trying to write 3 years ago was if you choose your actions, and one of those actions was to switch slot 1 to 3, but the opponent switches in Intimidate first, forcing Eject Pack to resolve and switching slot 2 to 3, that initial switch should still happen. So slot 1 would still switch to slot 3. Here's a PS replay of the buggy behavior to make it more clear (notice how we indicate the switch doesn't resolve): https://replay.pokemonshowdown.com/gen9doublescustomgame-2254947672

So in that test, it should have Wynaut on the left slot (battle.p2.active[0]) and Morelull on the right slot (battle.p2.active[1]).

There's another Eject Pack bug I don't have a test written for: https://www.smogon.com/forums/threads/various-eject-pack-bugs.3733517/post-10009741

Could you speak a bit more about your implementation? It's not clear to me why you're doing a Speed sort in runMoveEffects; when multiple Pokemon get hit by a spread move, e.g. Bulldoze, the order drops happen is ally -> left -> right, not based on Speed. Same thing for faints, same thing for the order Pokemon have their attack lowered from Intimidate, etc.

[ShivaD173]

Sorry, I haven't had time to work on this, and I saw that pyuk has started a PR doing some similar stuff, so if this just gets closed, thats fine by me. The speed sort was just wrong, it broke no other tests, so I thought it might work. This is done in a (slightly) better way now.

The main approach is waiting for all mons to "start" before a switch-out is called. In addition, it aggregates all switch-outs until the next time a switch out can be called, then does it by speed order. This does break the Eject Pack + eject button case, but this seems rare enough to be skipped until a proper refactor.