fix: deregister guardian from AllGuardians and slot index in resign_g…

[Joycejay17]

Summary

Closes #303 - resign_guardian now properly deregisters guardians from all membership structures (AllGuardians and the dense slot index), preventing permanent lockout of resigned guardians and eliminating storage leaks.

Problem

Previously, resign_guardian only cleared the guardian membership flag (DataKey::Guardian) but did not remove the address from:

• DataKey::AllGuardians set
• The dense slot index (GuardianIndexAt/GuardianIndexOf/GuardianIndexCount)

This caused:

1. Permanent lockout: A resigned guardian could never be re-added because add_guardian checks all_guardians.contains(guardian)
2. Storage leak: get_snapshot_meta().guardian_count remained inflated, and get_guardians_page still included resigned guardians
3. Inconsistency: resign_guardian and remove_guardian left the contract in different states for the same address

Solution

• Extracted a shared deregister_guardian helper from remove_guardian that handles complete membership cleanup
• Updated resign_guardian to call deregister_guardian instead of the bare remove(&g_key)
• Preserved existing token-refund and timelock logic that runs after deregistration
• Maintained the NotGuardian early return check

Changes

• src/guardian.rs: Added deregister_guardian helper function (pub(crate)), refactored remove_guardian to use it
• src/contracts/vault_ops.rs: Updated resign_guardian to call guardian::deregister_guardian
• tests/resign_guardian_deregister.rs: Added regression test test_resigned_guardian_can_be_readded

Test plan

• ✅ New regression test verifies resigned guardian can be re-added
• ✅ Test confirms guardian_count decreases by 1 after resignation
• ✅ Test confirms resigned guardian no longer appears in get_guardians_page
• ✅ Existing test test_resign_guardian_refunds_tokens still passes
• ✅ Contract tests pass: cargo test --lib
• ✅ New test passes: cargo test --test resign_guardian_deregister

Verification

After resignation:

• is_guardian(g) returns false
• get_snapshot_meta().guardian_count decreases by 1
• get_guardians_page(0, 50) no longer includes the resigned address
• add_guardian(admin, g) succeeds and the guardian is properly re-registered
• Both resign_guardian and remove_guardian now leave byte-identical guardian-index state for the same address