Tests: Test that a world can generate with item links

[alwaysintreble]

What is this fixing or adding?

Moves Item Link creation out of main; there's likely a better place but this is what made the most sense to me. Also adds a unit test for generating with full item link replacements. Since most worlds generate filler randomly, that filler won't actually get replaced, so unsure on how to best add actual tests beyond "doesn't crash".

[el-u]

I'm not sure how much value this actually adds since it isn't really testing anything beyond "it generates", but done.

Regarding added value:
The test now shows that there are a bunch of world types that only fail if replacement items are also linked. We already knew this for Messenger and Lufia II (see #2083), but for the other worlds that seems to be new and potentially useful information.

Regarding doing nothing beyond "it generates":
Would it make sense to assertBeatable?

[alwaysintreble]

Currently failing worlds:

• ALTTP: checking for a nonexistent "fake" option
• KH2: using self.random @JaredWeakStrike
• L2AC: using self.random
• Messenger: The Messenger: Hotfix item links filler gen #3078
• OOT: no attribute junk_ice_traps @espeon65536
• Pokemon RB: unable to place items @Alchav
• Super Metroid: unable to place items @lordlou
• SMZ3: no attribute junkItemsNames
• Stardew: using self.random @agilbert1412

all of the self.random issues are addressed in #2083, so for the relevant people this is just for notification and to confirm that 2083 solves those worlds. Messenger and L2AC have already been tested fixed with 2083.

[NewSoupVi]

The only remaining fails are SMZ3 and Pokemon RB.

A PR for SMZ3 has been opened: #4099
A PR for Pokemon RB has not.

Since it has been long enough, I will ping @Alchav again here. We really want to merge this unit test asap if possible and Pokemon RB is failing it.
Will probably ping in the Discord on Monday or so.