Add ONNX support for Embeddings and Pipelines

[davidmezzetti]

Add ONNX support for Embeddings and Pipelines.

Sequence to Sequence models (summarization, transcription, translation) will be added later once ONNX support for encoder-decoder models is more mature.

[nickchomey]

The HF documentation says that you can export seq2seq to ONNX with the OnnxSeq2SeqConfigWithPast class.
https://huggingface.co/docs/transformers/v4.23.1/en/main_classes/onnx#onnx-configurations

This was added with this PR in March huggingface/transformers#14700

Perhaps it is sufficient to be incorporated into txtai now? It would be great to be able to use ONNX versions of the various HF models, for their increased performance.

[davidmezzetti]

Haven't looked at this for quite a while, so it's possible. Would you mind opening a new issue?

[nickchomey]

#371 has been opened to track adding seq2seq export to ONNX