neo4j: Fix test for order-insensitive comparison and floating-point p…

…recision issues (#24338)

**Description:** 
This PR addresses two main issues in the `test_neo4jvector.py`:
1. **Order-insensitive Comparison:** Modified the
`test_retrieval_dictionary` to ensure that it passes regardless of the
order of returned values by parsing `page_content` into a structured
format (dictionary) before comparison.
2. **Floating-point Precision:** Updated
`test_neo4jvector_relevance_score` to handle minor floating-point
precision differences by using the `isclose` function for comparing
relevance scores with a relative tolerance.

Errors addressed:

- **test_neo4jvector_relevance_score:**
  ```
AssertionError: assert [(Document(page_content='foo', metadata={'page':
'0'}), 1.0000014305114746), (Document(page_content='bar',
metadata={'page': '1'}), 0.9998371005058289),
(Document(page_content='baz', metadata={'page': '2'}),
0.9993508458137512)] == [(Document(page_content='foo', metadata={'page':
'0'}), 1.0), (Document(page_content='bar', metadata={'page': '1'}),
0.9998376369476318), (Document(page_content='baz', metadata={'page':
'2'}), 0.9993523359298706)]
At index 0 diff: (Document(page_content='foo', metadata={'page': '0'}),
1.0000014305114746) != (Document(page_content='foo', metadata={'page':
'0'}), 1.0)
  Full diff:
  - [(Document(page_content='foo', metadata={'page': '0'}), 1.0),
+ [(Document(page_content='foo', metadata={'page': '0'}),
1.0000014305114746),
? +++++++++++++++
- (Document(page_content='bar', metadata={'page': '1'}),
0.9998376369476318),
? ^^^ ------
+ (Document(page_content='bar', metadata={'page': '1'}),
0.9998371005058289),
? ^^^^^^^^^
- (Document(page_content='baz', metadata={'page': '2'}),
0.9993523359298706),
? ----------
+ (Document(page_content='baz', metadata={'page': '2'}),
0.9993508458137512),
? ++++++++++
  ]
  ```

- **test_retrieval_dictionary:**
  ```
AssertionError: assert [Document(page_content='skills:\n- Python\n- Data
Analysis\n- Machine Learning\nname: John\nage: 30\n')] ==
[Document(page_content='skills:\n- Python\n- Data Analysis\n- Machine
Learning\nage: 30\nname: John\n')]
At index 0 diff: Document(page_content='skills:\n- Python\n- Data
Analysis\n- Machine Learning\nname: John\nage: 30\n') !=
Document(page_content='skills:\n- Python\n- Data Analysis\n- Machine
Learning\nage: 30\nname: John\n')
  Full diff:
- [Document(page_content='skills:\n- Python\n- Data Analysis\n- Machine
Learning\nage: 30\nname: John\n')]
? ---------
+ [Document(page_content='skills:\n- Python\n- Data Analysis\n- Machine
Learning\nage: John\nage: 30\n')]
? +++++++++
  ```

libs/community/tests/integration_tests/vectorstores/test_neo4jvector.py

@@ -1,9 +1,11 @@
 """Test Neo4jVector functionality."""
 
 import os
+from math import isclose
 from typing import Any, Dict, List, cast
 
 from langchain_core.documents import Document
+from yaml import safe_load
 
 from langchain_community.graphs import Neo4jGraph
 from langchain_community.vectorstores.neo4j_vector import (
@@ -217,12 +219,21 @@ def test_neo4jvector_relevance_score() -> None:
     )
 
     output = docsearch.similarity_search_with_relevance_scores("foo", k=3)
-    assert output == [
+    expected_output = [
         (Document(page_content="foo", metadata={"page": "0"}), 1.0),
         (Document(page_content="bar", metadata={"page": "1"}), 0.9998376369476318),
         (Document(page_content="baz", metadata={"page": "2"}), 0.9993523359298706),
     ]
 
+    # Check if the length of the outputs matches
+    assert len(output) == len(expected_output)
+
+    # Check if each document and its relevance score is close to the expected value
+    for (doc, score), (expected_doc, expected_score) in zip(output, expected_output):
+        assert doc.page_content == expected_doc.page_content
+        assert doc.metadata == expected_doc.metadata
+        assert isclose(score, expected_score, rel_tol=1e-5)
+
     drop_vector_indexes(docsearch)
 
 
@@ -779,8 +790,16 @@ def test_retrieval_dictionary() -> None:
             )
         )
     ]
+
     output = docsearch.similarity_search("Foo", k=1)
-    assert output == expected_output
+
+    def parse_document(doc: Document) -> Any:
+        return safe_load(doc.page_content)
+
+    parsed_expected = [parse_document(doc) for doc in expected_output]
+    parsed_output = [parse_document(doc) for doc in output]
+
+    assert parsed_output == parsed_expected
     drop_vector_indexes(docsearch)