llm 0.32.1 fixes fresh-install failures by temporarily pinning OpenAI dependency

The llm project released version 0.32.1 to fix a bug affecting fresh installations. The issue arose because the OpenAI Python library removed its use of httpx while llm relied on httpx directly but only received it transitively via the openai package, causing newly installed llm instances to fail. The 0.32.1 patch temporarily pins the openai package to openai<3 to restore functionality. The release specifies llm 0.32.1 and the recommended openai<3 constraint. The developers said this is only a temporary measure; the planned 0.33 release will switch from httpx to the httpx2 library as a longer-term solution. The change aims to improve compatibility and dependency stability and is particularly relevant for developers working with Python environments and administrators performing package installations. The release reduces downtime from disrupted fresh installations until the transition is finalized.