Add DALL-E: Zero-Shot Text-to-Image Generation

# 🚀 Feature request

Please add DALLE-E model to huggingface's Transformers library.

1. [Announcement](https://openai.com/blog/dall-e/)
2. [Abstract](https://arxiv.org/abs/2102.12092v2)
3. [Paper](https://arxiv.org/pdf/2102.12092v2.pdf)
4. Code:
- [openai/DALL-E](https://github.com/openai/DALL-E) (official)
- [lucidrains/DALLE-pytorch](https://github.com/lucidrains/DALLE-pytorch) ([Colab](https://colab.research.google.com/drive/1dWvA54k4fH8zAmiix3VXbg95uEIMfqQM?usp=sharing))

## Motivation

> DALL·E is a 12-billion parameter version of [GPT-3](https://huggingface.co/transformers/model_doc/gpt.html) trained to generate images from text descriptions, using a dataset of text–image pairs
>
> We (Open AI) decided to name our model using a portmanteau of the artist Salvador Dalí and Pixar’s WALL·E.