fix: embeddings registry edge cases and chunk-size accounting

Follow-up to the embeddings work, from a review pass over the branch.

- Route the OpenAI/Azure key handling through the model registry instead of
  matching the canonical name literally, so the `text-embedding-ada-002`
  alias the registry now accepts also reaches the Azure deployment name
  rather than failing every embed with DeploymentNotFound.
- Fall back to a default width where the embeddings model reports no
  dimension. A model outside the registry returns None rather than no
  attribute, so `getattr` with a default did not catch it and the width
  reached the DDL as `vector(None)` / `list_size=None`.
- Point HF_HUB_CACHE at the prefetch directory. Chunking loads the tokenizer
  through `tokenizers`, which reads the hub cache, so a fresh container
  fetched over the network on first ingest and an offline one silently fell
  back to cl100k.
- Charge a token that collapses a long unbroken run by its character span.
  WordPiece emits one [UNK] for any word over its character limit, which made
  base64 and minified content count as near-zero tokens, so nothing split it
  and oversized chunks reached the embedding server.
- Preserve chunk ids and honour --batch-size when rebuilding a FAISS index.
  Fresh uuids orphaned GraphRAG's graph_node_chunks rows, and the whole index
  went out in a single embed call on remote servers.
- Document that granite runs an int8-quantised graph, and scope the
  SentenceTransformer parity claim to mpnet's fp32 graph, which is where it
  was measured.
- Correct the embeddings docs: a matching dimension is not a matching model,
  so a same-width swap raises nothing and silently degrades retrieval.

tests/vectorstore/test_base.py

@@ -591,6 +591,49 @@ class TestGetEmbeddingsResolver:
             "openai_text-embedding-ada-002", model="embed-deploy"
         )
 
+    @patch("application.vectorstore.base.settings")
+    @patch("application.vectorstore.base.EmbeddingsSingleton.get_instance")
+    def test_openai_alias_also_reaches_the_azure_deployment(
+        self, mock_get_instance, mock_settings
+    ):
+        """The registry accepts the bare alias, so the key handling must too.
+
+        Matching on the canonical string alone sent the alias down the generic
+        branch, where the deployment name is never passed and Azure answers
+        every embed with DeploymentNotFound.
+        """
+        mock_settings.EMBEDDINGS_BASE_URL = None
+        mock_settings.OPENAI_API_BASE = "https://azure.openai.com"
+        mock_settings.OPENAI_API_VERSION = "2023-05-15"
+        mock_settings.AZURE_DEPLOYMENT_NAME = "deploy"
+        mock_settings.AZURE_EMBEDDINGS_DEPLOYMENT_NAME = "embed-deploy"
+        mock_settings.EMBEDDINGS_NAME = "text-embedding-ada-002"
+        mock_settings.EMBEDDINGS_KEY = "sk-key"
+
+        get_embeddings()
+
+        mock_get_instance.assert_called_once_with(
+            "text-embedding-ada-002", model="embed-deploy"
+        )
+
+    @patch("application.vectorstore.base.settings")
+    @patch("application.vectorstore.base.EmbeddingsSingleton.get_instance")
+    def test_openai_name_is_matched_case_insensitively(
+        self, mock_get_instance, mock_settings
+    ):
+        mock_settings.EMBEDDINGS_BASE_URL = None
+        mock_settings.OPENAI_API_BASE = None
+        mock_settings.OPENAI_API_VERSION = None
+        mock_settings.AZURE_DEPLOYMENT_NAME = None
+        mock_settings.EMBEDDINGS_NAME = "OpenAI_Text-Embedding-Ada-002"
+        mock_settings.EMBEDDINGS_KEY = "sk-from-settings"
+
+        get_embeddings()
+
+        mock_get_instance.assert_called_once_with(
+            "OpenAI_Text-Embedding-Ada-002", openai_api_key="sk-from-settings"
+        )
+
     @patch("application.vectorstore.base.settings")
     @patch("application.vectorstore.base.EmbeddingsSingleton.get_instance")
     def test_explicit_arguments_win_over_settings(