it: declare Article attribute unions and typed callers

[itsgrimetime]

Article.x4_specialAttributes hides different item layouts behind void*. Declare it as union ItemSpecialAttributes*, describe the concrete layouts in C, and access named members at item and fighter callers. Remove the DAT_TYPE override and attribute-pointer casts while preserving the original binary.

Reuse members for genuinely shared layouts, including Game & Watch/Kirby's Chef Pan and itSanseijuuAttributes for Entei/Raikou/Suicune. Give Freeze, Bunny Hood, Metal Box, and Kirby's dropped-copy ability their actual records; describe complete Pokémon layouts and food/animation/trailing arrays. Preserve established identifiers and unknown offset field names. The TODO update documents remaining discriminator and inline-array count support.

This general cleanup is independent of #3619 and uses the existing DAT CLI. Loader bindings, the fighter-slot union and required callers, and Samus's grapple sampling/count fix are in #3619. Neither PR requires the other to merge first. Remaining item-kind conditions and matrix typing/sampling are follow-ups.

Validation:

• The MWCC game build retains 100% matching code/data and passes the original DOL checksum.
• Clang syntax/AST checks pass for all 147 Article callers, with no attribute-pointer casts. Local commit checks pass.
• With the unchanged CLI, the Clang DAT pipeline matches 12,087/12,089 samples across 838 archives. The two existing mismatches are Samus's x8_hurtbones_x352D8 and Game & Watch's dyn_descs_0_x78F0.
• Rebased onto current master. Old Goomba uses master’s canonical itOldkuriAttributes declaration and typed nested fields through the oldkuri union member; there is no duplicate declaration. Its 16-byte layout is verified in DWARF. Original archives and generated output are excluded.

[ribbanya]

Are any of the dat-cli changes necessary still?

[itsgrimetime]

Are any of the dat-cli changes necessary still?

@ribbanya

yeah i'm pretty sure they're still needed for the samus fix (but not for the article cleanup). the rust changes are what brings the fighter kind from the loader's name table into the walk & letting DAT_IF select it in the right slot

without those changes it reports an ambiguous union at samus's slot 4

[ribbanya]

If it is necessary, please split the PRs. A src-only changeset would be much easier to review.