feat(fc2b): add SigLIP embedder wrapper

Direct port of ImageRepo's siglip.py. Lazy torch/transformers import so
the web container can import the module (for enqueue logic) without the
torch cost. EMBED_DIM=1152 asserted against the schema's Vector(1152)
columns. Real inference runs in the local integration suite.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

tests/test_ml_embedder.py

@@ -0,0 +1,30 @@
+"""Embedder unit tests. The SigLIP model is a multi-GB download not present
+in CI, so these test only the import-safety contract and the singleton.
+Real embedding runs in the local integration suite.
+"""
+
+from backend.app.services.ml.embedder import (
+    EMBED_DIM,
+    MODEL_VERSION,
+    Embedder,
+    get_embedder,
+)
+
+
+def test_embed_dim_matches_schema():
+    # image_record.siglip_embedding and tag_reference_embedding.embedding
+    # are both Vector(1152). This must stay in sync.
+    assert EMBED_DIM == 1152
+
+
+def test_model_version_default():
+    assert MODEL_VERSION == "siglip-so400m-patch14-384"
+
+
+def test_get_embedder_singleton():
+    assert get_embedder() is get_embedder()
+
+
+def test_not_loaded_until_called():
+    e = Embedder()
+    assert e._model is None  # not loaded until .load()