Simplify OpenAI VectorSearchRetrieverTool tool execution function

[leonbi100]

This PR refactors the OpenAI VectorSearchRetrieverTool#execute_tools function to be VectorSearchRetrieverTool#execute. The goal here is to do less of the response parsing and instead only focus on executing the retriever call, which is more in line with how the OpenAI SDK documents function calling.

The new usage pattern is more along the lines of:

dbvs_tool = VectorSearchRetrieverTool("index_name")
tools = [dbvs_tool.tool, ...]
response = openai.chat.completions.create(
  model="gpt-4o",
  messages=initial_messages,
  tools=tools,
)
tool_call = response.choices[0].message.tool_calls[0]
args = json.loads(tool_call.function.arguments)
retriever_resp = vector_search_tool.execute(query=args["query"])

[leonbi100]

Refactoring this to match how the UC Function Toolkit does this here

[smurching]

We might want to make this into a public API so that we can just officially recommend customers use it to get the name of the vector search tool, but ok to do that in a follow-up

[smurching]

LGTM!