fix: resolve data loss bugs in farm IP, node interface, and contract mapping handlers

[sameh-farouk]

Summary

Closes #208

Fixes 7 data corruption bugs in the event mapping handlers, discovered during a code audit.

Fixes

B1 — Race conditions in farmUpdated (farms.ts)
Replace async/await inside .forEach() with for...of loops for both IP save and IP removal iterations. Ensures sequential execution of read-modify-write operations.

B2 — PublicIp ID collision in farmStored (farms.ts)
Multiple public IPs in a FarmStored event all received the same ID (item.event.id), causing upsert overwrites. Now appends -{ip} to create unique IDs per IP.

B3 — Interfaces ID collision in nodeStored (nodes.ts)
Same collision pattern — multiple interfaces got the same ID. Now appends -{name}, matching the existing correct pattern in nodeUpdated.

B4 — Undefined contractID guard in grace period handlers (contracts.ts)
contractGracePeriodStarted and contractGracePeriodEnded had no guard when the event version was unrecognized. Added undefined check with error log and early return.

B5 — handlePublicConfigV105 fallthrough (nodes.ts)
When config was None (meaning "remove public config") and no existing record was found, the function fell through and created a ghost PublicConfig with all-null fields. Now returns early for the entire !config branch.

B6 — nodeStored never saved inline PublicConfig (nodes.ts)
In spec v9, create_node accepted a public_config parameter. getNodePublicConfig() extracted it but the result was never persisted. Now saves when non-null.

B7 — Cross-farm duplicate IP not actually skipped (farms.ts)
When an IP already existed on a different farm, the code logged "skipped" but continued to save — overwriting the original farm's gateway. Now properly continues after the error log. Also removed the no-op savedIP.ip reassignment.

Additional cleanup

• Added - separator to composite IDs for readability (eventId-eth0 instead of eventIdeth0)
• Applied separator consistently to both nodeStored and nodeUpdated interface IDs
• Fixed typo "skiped" → "skipped" in error log

Backwards compatibility

• All entity lookups use domain fields (ip, contractId, farm.id, node.nodeID), never the raw id — ID format changes are safe
• ctx.store.save() is an upsert (INSERT ... ON CONFLICT (id) DO UPDATE) — confirmed in Subsquid's @subsquid/typeorm-store
• Verified against tfchain pallet source: spec v9 (pallet 0.1.0-b1) accepted public_config in create_node; spec 11+ (pallet 0.1.0-b4) hardcoded None
• Chain does not validate IP uniqueness across farms (add_farm_ip only checks within same farm)

Test plan

• npx tsc --noEmit — compiles clean after each fix
• Create farm with multiple IPs on devnet → verify all IPs visible in GraphQL
• Create node with multiple interfaces → verify all interfaces visible
• Full re-index from genesis → compare results before and after

[sameh-farouk]

Total PublicIPs: Increased from 1,364 to 1,367 (+3)

• collision fixes

  • Farm 4432: Increased from 4 to 6 IPs (+2). We recovered 1.1.1.{1,2}/24, bringing the count in line with the 6 IPs found on-chain.

  • Farm 7944: Increased from 1 to 2 IPs (+1). We recovered 98.2.88.99/24. The test now matches the 2 IPs (98.2.88.99/24 and 98.2.88.89/24) found on-chain.

• cross-farm duplicate handling

  • Farm 4291: Decreased from 52 to 47 IPs (-5). It lost 125.25.25.{13,14,15}/16 (which belongs to farm 2586) and 1.1.1.{1,2}/24 (which belongs to farm 4432). These were previously present only because collision overwrites assigned IPs to whichever farm was processed last.

  • Farm 2586: Increased from 15 to 20 IPs (+5). We recovered 125.25.25.{11,12,13,14,15}/16. While there are 21 IPs on-chain, 125.25.25.25/16 was correctly skipped as it is already registered to farm 4376.

The fixed logic ensures the first farm keeps ownership while subsequent duplicates are logged and skipped. This match what was documnted, loged, and comunicated but it was broken.