Update `models.transformers` to use `SequenceGeneratorAdapter` and `OutlinesLogitsProcessors`

[lapp0]

In draft until https://github.com/huggingface/transformers/pull/31448 makes it into a new transformers release

Fixes #789

Addresses #806 (does everything except remove the issues requirement of "remove torch")

Closes https://github.com/outlines-dev/outlines/pull/910 (device inconsistency issue handled through other means)

Problem

• For #789 in SequenceGenerator if input_ids is empty, it fails.
• Generally, models.transformers has an inconsistent implementation with other modules making the codebase harder to manage.

Solution

• Use SequenceGeneratorAdapter for transformers instead of SequenceGenerator
• Implement Transformers.generate and Transformers.stream which use model.generate(logits_processor=...) argument with outlines.processors.OutlinesLogitsProcessor

Additional Changes

• Temporarily disables stop_at argument for transformers and implements test to determine whether upstream is fixed (https://github.com/huggingface/transformers/issues/31435)
  • fixed upstream, change reverted
• Update docs and examples to use seed instead of rng
• Disables outlines.generate.cfg for now (https://github.com/outlines-dev/outlines/issues/959)

TODO:

• [x] Determine cause of regression when generating json
  • was caused by GenerationConfig default max_tokens being 20
• [x] replace rng with seed in all tests and all documentation.
• [x] Awaiting transformers fix for https://github.com/huggingface/transformers/issues/31435 so stop_at works.