fix(purchases): resolve splits payout wallets historically

[rickyrombo]

Summary

The `v_usdc_purchases` view was joining the seller's current payout wallet (`users.spl_usdc_payout_wallet`). Anyone who changed payout wallets after a sale lost their `splits[*].user_id` attribution: the on-chain `sol_payments.to_account` (the wallet at purchase time) no longer equals the row's current `spl_usdc_payout_wallet`, so the LEFT JOIN produces NULL.

Two user-visible consequences:

• `/v1/users/{id}/purchases/download` shows `paid_to_artist: 0` because the route code matches splits to the seller by `split.user_id == sellerUserId`.
• `/v1/users/{id}/purchases` is slow — `users.spl_usdc_payout_wallet` isn't indexed, so the per-payment LEFT JOIN turns into a Seq Scan on `users` × splits × purchases × pages.

Fix

Replace the current-state join with a LATERAL lookup on `user_payout_wallet_history` filtered by `block_timestamp <= sp.created_at`. This mirrors Python's `add_wallet_info_to_splits` (`discovery-provider/src/queries/get_extended_purchase_gate.py:171`) which is what populated the legacy `usdc_purchases.splits` JSON.

Plus an index on `user_payout_wallet_history (spl_usdc_payout_wallet, block_timestamp)` — the PK is keyed `(user_id, block_timestamp)`, which is the wrong direction for our lookup (we know the wallet, want the user).

The fallback chain (`sol_claimable_accounts` → `users.wallet`) is unchanged. It still handles users who never set a custom payout wallet (their `to_account` is the default USDC user-bank PDA, which is stable over time).

Tests

New `api/v_usdc_purchases_splits_test.go` covers the four splits-resolution gotchas:

• Payout wallet changed after purchase — seller's old splits still attribute correctly via history.
• Multiple historical wallets — view picks the one current at purchase time, not the most recent.
• No payout wallet ever set — falls back to the `sol_claimable_accounts` chain.
• Unowned wallet (network share) — `splits[*].user_id` is null instead of an accidental match.

Incidental cleanups

• `database.seed`: added a `blocks` baseRow and put `"blocks"` first in the `entityTables` ordered list (`user_payout_wallet_history` has an FK on `blocks.number`, so without the explicit ordering the seed sometimes inserted history rows before their referenced block).
• Existing tests that previously set `users.spl_usdc_payout_wallet` to fake the old current-state lookup now also seed `user_payout_wallet_history` so they still resolve under the new join.

Test plan

• `go test ./api/ -run 'TestV1UsersPurchases|TestV1UsersSales|TestExploreBestSelling|TestVUsdcPurchasesSplits' -count=1` passes locally
• After deploy, hit the slow URL again — should be back to normal:
  `/v1/users/q7l8a/purchases?limit=50&offset=0&sort_direction=desc&sort_method=date`
• EXPLAIN ANALYZE shows Index Scan on `user_payout_wallet_history_wallet_idx` instead of Seq Scan on `users`

🤖 Generated with Claude Code