[Bug]: memmachine-configure's Qdrant default silently attaches to any Qdrant on localhost:6333, including another deployment's #1791

Describe the bug

When qdrant-client is installed (memmachine-server[qdrant], or uv sync --all-extras), memmachine-configure defaults the vector store to qdrant at localhost:6333. It never connects, which is by design (configuration_wizard.py:92-100), and it doesn't say which endpoint it chose. If another MemMachine deployment's Qdrant is already on that host, the new server silently writes into it. Both servers then share the same long_term_memory__registry collection, and the new server creates its own collections in the other deployment's store.

Steps to reproduce

1. Run a host with an existing Qdrant on localhost:6333, in use by another MemMachine.
2. uv sync --all-extras && uv run memmachine-configure, accepting the vector-store default.
3. uv run memmachine-server, then add one episode.
4. curl localhost:6333/collections: a new collection was created next to the other deployment's, and the registry is shared. The server logs only Qdrant client 'qdrant_vector_store' validated successfully.

Expected behavior

The wizard prints the Qdrant endpoint it will use and asks for confirmation. Better still, it detects an existing MemMachine registry there and warns. At minimum, the prompt should say that the default means "an already-running Qdrant on localhost:6333".

Environment

• MemMachine Version: 0.4.0-rc1 (ad8ff24)
• OS: Ubuntu, Python 3.12