feat: vectorize CAM retrieval with NumPy and add multi-worker support

- Replace scalar hamming distance with NumPy bitwise_count for batch retrieval
- Add ThreadPoolExecutor-based multi-worker query parallelism
- Improve missing dataset error message with generation command hint
- Increase DEFAULT_MAX_QUERIES from 128 to 8192 for meaningful throughput tests

tests/test_retrieval_benchmark_performance.py

@@ -1,20 +1,27 @@
 from __future__ import annotations
 
 import csv
+import importlib.util
 import sys
 from pathlib import Path
 
 
 HW_SIM_DIR = Path(__file__).resolve().parents[1] / "hw" / "sim"
-if str(HW_SIM_DIR) not in sys.path:
-    sys.path.insert(0, str(HW_SIM_DIR))
+if str(HW_SIM_DIR) in sys.path:
+    sys.path.remove(str(HW_SIM_DIR))
+sys.path.insert(0, str(HW_SIM_DIR))
+benchmark_path = HW_SIM_DIR / "benchmarks" / "retrieval" / "test_retrieval_benchmark.py"
+spec = importlib.util.spec_from_file_location("hw_retrieval_benchmark", benchmark_path)
+assert spec is not None
+assert spec.loader is not None
+hw_retrieval_benchmark = importlib.util.module_from_spec(spec)
+sys.modules[spec.name] = hw_retrieval_benchmark
+spec.loader.exec_module(hw_retrieval_benchmark)
 
-from benchmarks.retrieval.test_retrieval_benchmark import (  # noqa: E402
-    QueryTiming,
-    build_hardware_performance,
-    summarize_query_timings,
-    write_outputs,
-)
+QueryTiming = hw_retrieval_benchmark.QueryTiming
+build_hardware_performance = hw_retrieval_benchmark.build_hardware_performance
+summarize_query_timings = hw_retrieval_benchmark.summarize_query_timings
+write_outputs = hw_retrieval_benchmark.write_outputs
 
 
 def test_summarize_query_timings_uses_query_only_accept_to_last_cycles() -> None: