support TextDiffuser pipelines natively in Diffusers

### Model/Pipeline/Scheduler description


TextDiffuser is an open-source framework and pretrained model series for generating unique depth maps that can be used to guide the model for typography.

It not only has open code and weights, but the dataset is also available, containing ~9.7 million typography samples 🤯

### Open source status

- [X] The model implementation is available.
- [X] The model weights are available (Only relevant if addition is not a scheduler).

### Provide useful links for the implementation

Original repository: https://github.com/microsoft/unilm/tree/master/textdiffuser
Follow-up repository: https://github.com/microsoft/unilm/tree/master/textdiffuser-2
Demo video (not mine): https://youtu.be/ApcJ1UyLQB8?t=62 (timestamp)

Start by reviewing the TextDiffuser and TextDiffuser-2 repositories linked in the issue to understand the available model implementations and weights. Then inspect Diffusers' existing pipeline integration patterns. Done means TextDiffuser pipelines are supported natively in Diffusers, with the required upstream functionality and validation identified during implementation.

Vom Indexierungsmodell aus dem Issue-Text verfasst.