add a benchmark for connection time versus number of tables

SQLite performs poorly for connections as the number of tables increase.

Do we perform better? the same? worse?

Right now the answer is worse. Add a benchmark to help us compare.

perf/connection/plot.py

@@ -0,0 +1,55 @@
+#!/usr/bin/env python3
+"""Plot connection benchmark results from CSV data."""
+
+import csv
+import sys
+
+import matplotlib.pyplot as plt
+
+
+def main() -> None:
+    """Plot benchmark results from CSV file."""
+    if len(sys.argv) != 2:
+        print("Usage: python3 plot.py <results.csv>")
+        sys.exit(1)
+
+    # Read the CSV file
+    databases = []
+    p50 = []
+    p95 = []
+    p99 = []
+
+    with open(sys.argv[1]) as file:
+        reader = csv.DictReader(file)
+        for row in reader:
+            databases.append(row["database"])
+            p50.append(int(row["p50"]))
+            p95.append(int(row["p95"]))
+            p99.append(int(row["p99"]))
+
+    x = range(len(databases))
+    width = 0.25
+
+    plt.figure(figsize=(12, 8))
+    plt.bar([i - width for i in x], p50, width, label="p50", alpha=0.8)
+    plt.bar([i for i in x], p95, width, label="p95", alpha=0.8)
+    plt.bar([i + width for i in x], p99, width, label="p99", alpha=0.8)
+
+    plt.xlabel("Database (Number of Tables)")
+    plt.ylabel("Connection Time (nanoseconds)")
+    plt.title("Database Connection Performance by Table Count")
+    plt.xticks(x, databases)
+    plt.legend()
+    plt.yscale("log")  # Use log scale for better visualization
+    plt.grid(True, alpha=0.3)
+
+    plt.tight_layout()
+    plt.savefig("connection_benchmark.png", dpi=300, bbox_inches="tight")
+    plt.show()
+
+    print("Chart saved as connection_benchmark.png")
+
+
+if __name__ == "__main__":
+    main()
+