fix(request): stop Request.from_url mutating caller user_data and writing empty __crawlee

[anxkhn]

Request.from_url mutates the caller's user_data dict in place, and it always
writes an empty __crawlee block even when no crawlee options are supplied. Both
were introduced by #1603.

• It took a live reference to the caller's nested block
  (crawlee_data_dict = user_data_dict.get('__crawlee', {})), then mutated it in
  place (crawlee_data_dict['maxRetries' | 'enqueueStrategy'] = ...) and replaced
  the __crawlee key on the caller's dict with a CrawleeRequestData model.
  Reusing a user_data dict therefore leaks enqueueStrategy / maxRetries into
  the caller's nested dict, and passing the same dict object twice raises
  TypeError: 'CrawleeRequestData' object does not support item assignment on the
  second call (the key is no longer a plain dict). The framework's own
  enqueue_links dodges the crash by building a fresh outer dict per link
  ({**base_user_data}), but the shared nested __crawlee dict still accumulates
  enqueueStrategy across links.
• The if crawlee_data: guard, meant to skip writing __crawlee when there is no
  crawlee data, is always true because an empty pydantic model is truthy
  (bool(CrawleeRequestData()) is True). So Request.from_url('https://example.com')
  writes user_data = {'__crawlee': {}} instead of leaving user_data empty.

Fix: copy both the outer user_data dict and the nested __crawlee block before
mutating them, and compare against CrawleeRequestData() so the write is skipped
when empty.

-        user_data_dict = kwargs.pop('user_data', {}) or {}
-        crawlee_data_dict = user_data_dict.get('__crawlee', {})
+        user_data_dict = dict(kwargs.pop('user_data', {}) or {})
+        crawlee_data_dict = dict(user_data_dict.get('__crawlee') or {})

         if max_retries is not None:
             crawlee_data_dict['maxRetries'] = max_retries
@@
         crawlee_data = CrawleeRequestData(**crawlee_data_dict)

-        if crawlee_data:
+        if crawlee_data != CrawleeRequestData():
             user_data_dict['__crawlee'] = crawlee_data

Issues

• No existing GitHub issue. This is a self-identified correctness bug found while
  reviewing Request.from_url. Happy to open a tracking issue with the repro
  first if you'd prefer that.

Testing

• Added tests/unit/test_request.py with four tests:
  • test_from_url_does_not_mutate_caller_user_data: the caller's dict and its
    nested __crawlee block are left untouched (same object, unchanged value).
  • test_from_url_can_reuse_same_user_data_dict: calling from_url twice with
    the same user_data object no longer raises TypeError.
  • test_from_url_omits_crawlee_data_when_empty: a bare from_url does not write
    __crawlee into user_data.
  • test_from_url_keeps_crawlee_data_when_supplied: the positive round-trip,
    max_retries / enqueue_strategy are still serialized into __crawlee and
    reflected on the request.
• Verified fail-first: reverting only src/crawlee/_request.py (keeping the new
  tests) fails 3 of the 4 tests, each on its targeted defect; all 4 pass with the
  fix.
• uv run pytest tests/unit/test_request.py -> 4 passed.
• Regression: uv run pytest tests/unit/crawlers/_basic/test_basic_crawler.py tests/unit/storages/test_request_queue.py tests/unit/test_router.py and the
  enqueue_strategy / max_retries cases -> all pass.
• uv run poe lint (ruff format + check) passes on the changed files.

Checklist

• CI passed