Deprecate low use models (#30781)

* Deprecate models
- graphormer
- time_series_transformer
- xlm_prophetnet
- qdqbert
- nat
- ernie_m
- tvlt
- nezha
- mega
- jukebox
- vit_hybrid
- x_clip
- deta
- speech_to_text_2
- efficientformer
- realm
- gptsan_japanese

* Fix up

* Fix speech2text2 imports

* Make sure message isn't indented

* Fix docstrings

* Correctly map for deprecated models from model_type

* Uncomment out

* Add back time series transformer and x-clip

* Import fix and fix-up

* Fix up with updated ruff

docs/source/en/model_doc/realm.md

@@ -16,6 +16,14 @@ rendered properly in your Markdown viewer.
 
 # REALM
 
+<Tip warning={true}>
+
+This model is in maintenance mode only, we don't accept any new PRs changing its code.
+If you run into any issues running this model, please reinstall the last version that supported this model: v4.40.2.
+You can do so by running the following command: `pip install -U transformers==4.40.2`.
+
+</Tip>
+
 ## Overview
 
 The REALM model was proposed in [REALM: Retrieval-Augmented Language Model Pre-Training](https://arxiv.org/abs/2002.08909) by Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat and Ming-Wei Chang. It's a
@@ -86,4 +94,4 @@ This model was contributed by [qqaatw](https://huggingface.co/qqaatw). The origi
 
 [[autodoc]] RealmForOpenQA
     - block_embedding_to
-    - forward
+    - forward