Add support for HubertForCTC

[jimypbr]

What does this PR do?

Adds support for HubertForCTC model.

Can fine-tune it on librispeech with the command:

python examples/speech-recognition/run_speech_recognition_ctc.py \
    --dataset_name "librispeech_asr" \
    --dataset_config_name "clean" \
    --train_split_name "train.100" \
    --eval_split_name "validation" \
    --model_name_or_path facebook/hubert-base-ls960 \
    --ipu_config_name Graphcore/wav2vec2-ctc-base-ipu \
    --ipu_config_overrides "device_iterations=16,inference_device_iterations=16" \
    --mask_time_prob 0.0 \
    --mask_feature_prob 0.0 \
    --output_dir "./hubert-base-960h" \
    --overwrite_output_dir \
    --length_column_name "input_length" \
    --num_train_epochs 5 \
    --learning_rate "3e-4" \
    --warmup_steps 400 \
    --evaluation_strategy "steps" \
    --text_column_name "text" \
    --save_steps 400 \
    --eval_steps 400 \
    --logging_steps 10 \
    --save_total_limit 1 \
    --freeze_feature_encoder \
    --do_train \
    --do_eval \
    --layerdrop 0.0 \
    --per_device_train_batch_size 1 \
    --per_device_eval_batch_size 2 \
    --pod_type pod8 \
    --adam_beta1 0.9 \
    --adam_beta2 0.98 \
    --adam_epsilon 0.0001 \
    --report_to wandb \
    --dataloader_drop_last \
    --dataloader_mode "async_rebatched" \
    --dataloader_num_workers 8

Fixes # (issue)

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[katalinic-gc]

in parallelize the feature encoder gets frozen; does any "unfreezing" need to happen here?