feat: integrate Qdrant vector embeddings via sentence-transformers (Phase 1 Step 3)

[vineethwilson15]

Goal

Wire sentence-transformers into the indexer so every file produces a vector in Qdrant.

Acceptance Criteria

• EmbeddingClient wrapping SentenceTransformer(EMBED_MODEL)
• EMBED_MODEL env var (default all-MiniLM-L6-v2)
• Qdrant collection auto-created on first run
• _write_atomically() embeds + upserts to Qdrant atomically with Neo4j write
• Unit tests mock EmbeddingClient