fix(parser/email): align metadata/text_html emission with Python flow contract (#18005)

Aligns the Go EML parser (`internal/parser/parser/email_parser.go`) with the Python flow parser (`rag/flow/parser/parser.py:_email`) on two structural points so the Go path is a faithful drop-in for the Python path on `.eml` inputs.

deepdoc/parser/html_parser.py

@@ -233,6 +233,13 @@ class RAGFlowHtmlParser:
     def _token_count(cls, text):
         if not text:
             return 0
+        # FIXME: The identity of the tokenizer behind `rag_tokenizer.tokenize`
+        # (used here to size HTML chunks) is NOT confirmed. The earlier assumption
+        # that it resolves to `infinity.rag_tokenizer.RagTokenizer` is incorrect.
+        # Verify what `rag_tokenizer.tokenize` actually resolves to before claiming
+        # Parser and TokenChunker use different tokenizers. For reference,
+        # TokenChunker uses `num_tokens_from_string` (common/token_utils.py) which
+        # is tiktoken cl100k_base. See PARSER_ALIGNMENT_HANDOFF.md §3.2.
         tks_str = rag_tokenizer.tokenize(text)
         return len(tks_str.split(" ")) if tks_str else 0