Revamp questionnaire, parallelize run-all, add new tasks

- Replace 6 compound Likert questions with 12 atomic ones grouped by
  dimension (syntax, expressiveness, data/IO, errors, overall); drop
  free-form question. Responses now stored as ints, not strings.
- Back-compat layer maps legacy keys to new dimensions so existing
  results still render.
- Parallelize run-all with ThreadPoolExecutor (configurable workers)
  and add a thread-safe min-request-interval rate limiter to the
  Anthropic provider.
- Add new tasks: path_normalizer, todo_manager, currency_converter,
  locale_weather_url, network_info_parser, url_normalizer.

lush_bench/providers/anthropic.py

@@ -1,6 +1,8 @@
 from __future__ import annotations
 
 import os
+import threading
+import time
 from typing import Any
 
 import anthropic
@@ -17,8 +19,17 @@ class AnthropicProvider:
         self._client = anthropic.Anthropic(api_key=api_key)
         self._model = config.get("model", "claude-sonnet-4-20250514")
         self._max_tokens = config.get("max_tokens", 4096)
+        self._min_request_interval = config.get("min_request_interval", 0.1)
+        self._last_request_time = 0.0
+        self._lock = threading.Lock()
 
     def send(self, messages: list[Message], system: str = "") -> str:
+        with self._lock:
+            elapsed = time.monotonic() - self._last_request_time
+            if elapsed < self._min_request_interval:
+                time.sleep(self._min_request_interval - elapsed)
+            self._last_request_time = time.monotonic()
+
         api_messages = [{"role": m.role, "content": m.content} for m in messages]
         kwargs: dict[str, Any] = {
             "model": self._model,