feat(ingester): add CAP-46 NFT indexing with metadata caching

[Ipramking]

Summary

• NFT auto-detection via ABI sniff: events with 4 topics [Symbol(transfer), Address(from), Address(to), ScVal(token_id)] are classified as CAP-46 NFT transfers; SEP-41 fungible transfers (3 topics + i128 value) continue on the existing path unchanged
• New src/ingester/nft.ts: isNftTransferEvent, parseNftTransferEvent, parseNftEvents, fetchNftMetadata (lazy simulation via token_uri / name contract calls)
• Two new Prisma models: NftTransfer + NftMetadata with appropriate indexes
• New DB helpers: upsertNftTransfers, queryNftTransfers, getNftOwner, upsertNftMetadata, getNftMetadata
• Indexer splits events by topic count, processes both paths, and lazy-loads metadata per unique (contractId, tokenId) on first sight
• Two new REST endpoints:
  • GET /nfts/transfers?contract=…&token_id=…&address=…
  • GET /nfts/owners/:contract/:token_id — returns current owner + cached metadata
• NFT_CONTRACT_IDS env var for explicit watch list (auto-detection works without it)
• 12 unit tests in src/__tests__/nft.test.ts

Test plan

• npm test passes — NFT parser unit tests green
• Set NFT_CONTRACT_IDS=<testnet contract> and confirm NFT transfers populate nft_transfers table
• GET /nfts/transfers?contract=<id> returns paginated results
• GET /nfts/transfers?contract=<id>&token_id=<id> filters correctly
• GET /nfts/owners/<contract>/<token_id> returns owner + metadata
• GET /nfts/owners/<contract>/<unknown> returns 404
• Existing /transfers/* endpoints unaffected

Closes #58

[drips-wave]

@Ipramking Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits

[Miracle656]

Clean CAP-46 NFT indexer. Highlights:

• Detection heuristic (4 topics + Symbol("transfer") + first two scvAddress) is exactly what CAP-46 specifies. Fungible (3-topic + i128 value) keeps flowing through the existing path untouched — no risk of regression.
• Token ID decoding handles u128 → decimal string, Uint8Array → hex, string passthrough, with an XDR base64 fallback so no event is ever lost.
• Prisma indexes are right: contractId, tokenId, toAddress, fromAddress, plus the compound (contractId, tokenId) covers the owner-lookup path.
• NftMetadata cache with @@unique([contractId, tokenId]) + lazy fetch via fetchNftMetadata calling token_uri / name is the standard memoization shape — won't hammer the RPC.
• 12 unit tests cover happy paths and edge cases: non-transfer symbols, empty topics, wrong ScVal types, large u128 (2n**64n + 99n). Fixtures use nativeToScVal + Address.fromString().toScVal() correctly.
• NFT_CONTRACT_IDS env as an explicit watch list (auto-detection still works without it) gives operators a per-deployment escape hatch.

Two new endpoints (GET /nfts/transfers and GET /nfts/owners/:contract/:token_id) follow the existing route patterns.

Merging. Closes #58.