unCLIP variant (#2297)

* pipeline_variant

* Add docs for when clip_stats_path is specified

* Update src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* Update src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* Update src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* Update src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* prepare_latents # Copied from re: @patrickvonplaten

* NoiseAugmentor->ImageNormalizer

* stable_unclip_prior default to None re: @patrickvonplaten

* prepare_prior_extra_step_kwargs

* prior denoising scale model input

* {DDIM,DDPM}Scheduler -> KarrasDiffusionSchedulers re: @patrickvonplaten

* docs

* Update docs/source/en/api/pipelines/stable_unclip.mdx
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

---------
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

tests/test_pipelines_common.py

@@ -486,6 +486,9 @@ class PipelineTesterMixin:
         reason="XFormers attention is only available with CUDA and `xformers` installed",
     )
     def test_xformers_attention_forwardGenerator_pass(self):
+        self._test_xformers_attention_forwardGenerator_pass()
+
+    def _test_xformers_attention_forwardGenerator_pass(self, test_max_difference=True):
         if not self.test_xformers_attention:
             return
 
@@ -501,9 +504,12 @@ class PipelineTesterMixin:
         inputs = self.get_dummy_inputs(torch_device)
         output_with_offload = pipe(**inputs)[0]
 
+        if test_max_difference:
             max_diff = np.abs(output_with_offload - output_without_offload).max()
             self.assertLess(max_diff, 1e-4, "XFormers attention should not affect the inference results")
 
+        assert_mean_pixel_difference(output_with_offload[0], output_without_offload[0])
+
     def test_progress_bar(self):
         components = self.get_dummy_components()
         pipe = self.pipeline_class(**components)