Add opportunity detection and storage functionality with async processing

- Introduced OpportunityEvent class for structured opportunity data.
- Enhanced IncrementalCycleDetector to generate opportunities based on updated pairs.
- Implemented AsyncOpportunityWriter for persisting opportunities to the database.
- Updated MarketDataFeed to handle opportunity detection and execution in both paper and live trading modes.
- Added unit tests for opportunity detection and persistence.

src/arbitrade/storage/repositories.py

@@ -18,6 +18,16 @@ class MarketSnapshotRecord:
     latency_ms: float | None
 
 
+@dataclass(slots=True)
+class OpportunityRecord:
+    detected_at: datetime
+    cycle: str
+    gross_pct: float
+    net_pct: float
+    est_profit: float
+    executed: bool = False
+
+
 class MarketSnapshotRepository:
     def __init__(self, store: DuckDBStore) -> None:
         self._store = store
@@ -37,3 +47,32 @@ class MarketSnapshotRepository:
                     record.latency_ms,
                 ],
             )
+
+
+class OpportunityRepository:
+    def __init__(self, store: DuckDBStore) -> None:
+        self._store = store
+
+    def insert(self, record: OpportunityRecord) -> None:
+        with self._store.connect() as conn:
+            conn.execute(
+                """
+                INSERT INTO opportunities (
+                    detected_at,
+                    cycle,
+                    gross_pct,
+                    net_pct,
+                    est_profit,
+                    executed
+                )
+                VALUES (?, ?, ?, ?, ?, ?)
+                """,
+                [
+                    record.detected_at,
+                    record.cycle,
+                    record.gross_pct,
+                    record.net_pct,
+                    record.est_profit,
+                    record.executed,
+                ],
+            )