Fix seq2seq collator padding (#30556)

* fix seq2seq data collator to respect the given padding strategy

further added tests for the seq2seq data collator in the style of the `data_collator_for_token_classification` (pt, tf, np)

* formatting and change bool equals "==" to "is"

* add missed return types in tests

* update numpy test as it can handle unequal shapes, not like pt or tf

examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py

@@ -122,7 +122,8 @@ class ModelArguments:
         metadata={"help": "Deprecated. Please use the `language` and `task` arguments instead."},
     )
     suppress_tokens: List[int] = field(
-        default=None, metadata={
+        default=None,
+        metadata={
             "help": (
                 "Deprecated. The use of `suppress_tokens` should not be required for the majority of fine-tuning examples."
                 "Should you need to use `suppress_tokens`, please manually update them in the fine-tuning script directly."