fix: collector WebSocket trade parsing + DuckDB 1.5 API compatibility

data/collectors/hyperliquid.py:
- Fixed _handle_trades to handle HL WebSocket trade data as list[dict]
  (each trade dict has its own 'coin' field) instead of assuming single dict
- Fixed _poll_funding and _poll_open_interest to guard against
  data[0] being list vs dict (mainnet vs testnet response format difference)

data/duckdb_load.py:
- Replaced deprecated duckdb.from_sequence() with executemany() for
  l2_snapshots, trades, and funding table inserts (DuckDB 1.5.x API)

Tick runner verified on real mainnet BTC data:
- 324 events (45 L2 + 279 trades), VPIN-gated A-S MM, 1 trade, 98 cancels
- HFT dashboard generated with 8 panels from real tick data

data/duckdb_load.py

@@ -197,10 +197,9 @@ class DuckDBLoader:
             ))
 
         if rows:
-            import duckdb
-            rel = duckdb.from_sequence(rows)
-            self.conn.execute(
-                "INSERT OR IGNORE INTO l2_snapshots SELECT * FROM rel"
+            self.conn.executemany(
+                "INSERT OR IGNORE INTO l2_snapshots VALUES (?,?,?,?,?,?,?,?,?,?,?,?,?)",
+                rows
             )
             logger.info("Loaded %d L2 snapshots for %s", len(rows), coin)
 
@@ -242,10 +241,9 @@ class DuckDBLoader:
             ))
 
         if rows:
-            import duckdb
-            rel = duckdb.from_sequence(rows)
-            self.conn.execute(
-                "INSERT INTO trades SELECT * FROM rel"
+            self.conn.executemany(
+                "INSERT INTO trades VALUES (?,?,?,?,?,?,?)",
+                rows
             )
             logger.info("Loaded %d trades for %s", len(rows), coin)
 
@@ -282,10 +280,9 @@ class DuckDBLoader:
             ))
 
         if rows:
-            import duckdb
-            rel = duckdb.from_sequence(rows)
-            self.conn.execute(
-                "INSERT OR IGNORE INTO funding SELECT * FROM rel"
+            self.conn.executemany(
+                "INSERT OR IGNORE INTO funding VALUES (?,?,?,?,?,?)",
+                rows
             )
             logger.info("Loaded %d funding observations for %s", len(rows), coin)