fix: set resolve_entities=False in partition_xml (#3088)

### Summary

Closes #3078. Sets `resolve_entities=False` for parsing XML with `lxml`
in `partition_xml` to avoid text being dynamically injected into the
document.

### Testing

`pytest test_unstructured/partition/test_xml.py` continues to pass with
the update.

CHANGELOG.md

@@ -1,4 +1,4 @@
-## 0.14.3-dev1
+## 0.14.3-dev2
 
 ### Enhancements
 
@@ -8,6 +8,8 @@
 
 ### Fixes
 
+**Turn off XML resolve entities** Sets `resolve_entities=False` for XML parsing with `lxml`
+  to avoid text being dynamically injected into the XML document.
 * Add the missing `form_extraction_skip_tables` argument to the `partition_pdf_or_image` call.
 
 ## 0.14.2

unstructured/__version__.py

@@ -1 +1 @@
-__version__ = "0.14.3-dev1"  # pragma: no cover
+__version__ = "0.14.3-dev2"  # pragma: no cover

unstructured/partition/xml.py

@@ -51,7 +51,7 @@ def _get_leaf_elements(
     """Parse the XML tree in a memory efficient manner if possible."""
     element_stack = []
 
-    element_iterator = etree.iterparse(file, events=("start", "end"))
+    element_iterator = etree.iterparse(file, events=("start", "end"), resolve_entities=False)
     # NOTE(alan) If xml_path is used for filtering, I've yet to find a good way to stream
     # elements through in a memory efficient way, so we bite the bullet and load it all into
     # memory.