TST: Refactor remaining common tests to use pytest

[BenjaminBossan]

Follow up to #2462 and #2478

This deals with all the remaining tests that rely/relied on PeftCommonTester. For each test file, I created a separate commit, which may help reviewing.

After this is merged, I'll probably do another PR to do some clean up (e.g. no longer using self.skipTest, eliminating code duplication).

This is what I had to say to each file:

Refactor test_adaption_prompt.py

• Did not really use PeftCommonTester, thus removed it
• Removed skip if llama or mistral not available (this would only be necessary for very old transformers versions)
• Parametrized tests instead of duplicating
• Use small models from Hub instead of creating new ones
• Test coverage misses 3 more lines around loading checkpoint, but those seem unrelated to adaption prompt but instead due to using hub models instead of creating new ones

Refactor test_multitask_prompt_tuning.py

Same arguments apply as for test_adaption_prompt.py

Refactor test_feature_extraction.py

Pretty straightforward, test coverage is 100% identical.

Refactor test_stablediffusion.py

This was also pretty straightforward. After refactoring, the test coverage was 100% the same.

I noticed, however, that these tests did not cover LoKr, they only pretended to:

peft/tests/test_stablediffusion.py

Lines 113 to 114 in 37f8dc3

	"loha": (LoHaConfig, CONFIG_TESTING_KWARGS[1]),
	"lokr": (LoHaConfig, CONFIG_TESTING_KWARGS[1]),

Thus I added LoKr to the test matrix, after which the test coverage if of course different, but is fine. Unfortunately, LoKr merging tests fail on CPU because the outputs change after merging. This is unrelated to the PR, so let's just skip those tests for now.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.