feat(agent): scrape linked pages to ground answers in real content

Adds `_fetch_page_content()`, which downloads a URL with a
browser-like User-Agent and uses BeautifulSoup with the lxml
parser to strip scripts, navigation, and other non-content
markup, returning up to 3000 characters of readable text.

The response validator now also compares generated replies
against this scraped text, flagging named entities and claims
that do not appear in the retrieved page content.

Dependencies: adds beautifulsoup4==4.12.3 and pins lxml==5.3.0
for parser compatibility.

venv/lib/python3.11/site-packages/lxml/_elementpath.py

@@ -1,4 +1,4 @@
-# cython: language_level=3
+# cython: language_level=2
 
 #
 # ElementTree
@@ -85,8 +85,6 @@ def xpath_tokenizer(pattern, namespaces=None, with_prefixes=True):
                     yield ttype, "{%s}%s" % (namespaces[prefix], uri)
                 except KeyError:
                     raise SyntaxError("prefix %r not found in prefix map" % prefix)
-            elif tag.isdecimal():
-                yield token  # index
             elif default_namespace and not parsing_attribute:
                 yield ttype, "{%s}%s" % (default_namespace, tag)
             else: