Add benchmark for saved state performance

- Introduced `benchmark_saved_state.gd` to evaluate the performance of the saved state restoration process.
- Implemented JSON integer restoration to handle large integers accurately.
- Added command-line argument parsing for specifying the path to the diagnostic JSON file and optional delta seconds.
- Conducted performance tests at different speeds and logged the elapsed time for six consecutive steps.
- Created a new UID file for the benchmark script.

tests/test_short_lived_events.gd

@@ -75,5 +75,20 @@ func run() -> void:
     for batch in fast.state.batches.SN: total = E.add(total, batch.amount)
     check(E.eq(total, 10000), "Direct completions plus survivors conserve event count")
     dispose(fast)
+    # Sparse explosions: thousands of time windows but very few actual events.
+    # Exact interval edges and partial final windows exercise the jump boundary.
+    for count in [1, 3, 17]:
+        for year in [500, 1001]:
+            fast = fixture(false)
+            var legacy = fixture(true)
+            fast.state.amount.RHG = count
+            fast.state.batches.RHG = [{"amount": count, "start_year": E.of("0.25"), "end_year": E.of("1000.25")}]
+            legacy.state = fast.state.duplicate(true)
+            fast.evolution.evolve_batches(fast.state, year)
+            legacy.evolution.evolve_batches(legacy.state, year)
+            for field in fast.state:
+                check(equivalent(fast.state[field], legacy.state[field]), "Sparse %s at %s: %s" % [count, year, field])
+            dispose(fast)
+            dispose(legacy)
     print(("Short-lived events: %d checks, %d failures" % [checks, failures]))
     quit((0 if E.eq(failures, 0) else 1))