Fix autobonus not working with ammo, item combos, and pets

[MrKeiKun]

Pull Request Prelude

• I have followed proper Hercules code styling.
• I have read and understood the contribution guidelines before making this PR.
• I am aware that this PR may be closed if the above-mentioned criteria are not fulfilled.

Changes Proposed

autobonus()/autobonus2()/autobonus3() identified the source of an autobonus purely by reading status->current_equip_item_index (an inventory index) and deriving an EQP_* position bitmask from it. That global was only correctly maintained by the plain equip-item loop and the card loop in status_calc_pc_(), so every other script-execution path silently failed to register a working autobonus:

• Ammo: the ammo script-execution block never set status->current_equip_item_index, so autobonus() called from an ammo item's script either no-op'd or registered against whatever equip slot was last processed by the main loop.
• Item combos: combo scripts have no single backing equip slot, so there was no way to give them a valid position at all.
• Pets: the pet equip-script block runs after current_equip_item_index is explicitly reset to -1, so autobonus() there was a guaranteed no-op.
• Item + card collision: an item's own script and a card socketed in it both run with current_equip_item_index pointing at the same inventory slot, so their autobonuses shared the exact same identity and could suppress or interfere with each other.

This PR replaces the single EQP_*-bitmask identity with a proper struct s_autobonus_source { pos, card_id, combo_pos, pet_id }, threaded through pc_addautobonus/pc_delautobonus/pc_exeautobonus/autobonus_is_active and the three autobonus*() builtins. New context globals (current_equip_combo_pos, current_equip_pet_id) mirror the existing current_equip_card_id pattern and are set around the combo and pet-equip script executions in status_calc_pc_(). The old sd->state.autobonus position bitmask (which could only represent one active autobonus per equip slot) is removed in favor of per-entry lookups that re-validate the exact source (equip slot + specific card, or specific combo id, or specific pet id) against the character's live state.

Also fixes a related gap: pc_unequipitem() used to clear autobonus activation via that same bitmask unconditionally on unequip; since status_calc_pc() isn't always called synchronously after an unequip (e.g. PCUNEQUIPITEM_FORCE without PCUNEQUIPITEM_RECALC), this PR now calls pc->delautobonus(..., true) directly in pc_unequipitem() so an unequipped item's autobonus timer is always stopped promptly regardless of the caller's recalc flag.

Regenerated the affected HPMHooking trampolines/typedefs by hand for the changed pc->addautobonus/new pc->autobonus_is_active interface entries, since pc_addautobonus's signature changed; maintainers may want to re-run HPMHookGen.pl to confirm these match exactly.

Issues addressed: #2764