fix(embedder): or-chain on numpy array raises 'truth value ambiguous' ValueError -- crashes every BgeM3 embed call #287

Summary

BgeM3Embedder.embed() at line 232 uses: dense = result.get("dense_vecs") or result.get("dense") or result.get("embeddings"). When FlagEmbedding returns dense_vecs as a non-empty 2D numpy array (which it does on success), Python evaluates bool(numpy_2d_array) for the or short-circuit, raising ValueError: The truth value of an array with more than one element is ambiguous. This makes every successful embed call crash.

Root cause

Fix #245 added the or-chain to probe multiple output keys. The or operator requires bool() evaluation which is ambiguous for multi-element numpy arrays.

Fix

ingestion/embedder.py line 232: use explicit None check instead of or:
dense = next((v for k in ("dense_vecs", "dense", "embeddings") if (v := result.get(k)) is not None), None)

Impact

Every document ingestion using BgeM3Embedder crashes with ValueError when FlagEmbedding returns numpy arrays (which is the default return type).