[wwb]: load transformers model first, then only trust_remote_code

[eaidova]

General huggingface recommendation is to use trust_remote_code only if it is necessary, andthe model requires it, as potentially remote code may try to run arbitrary insecure code.
Besides that some models (e.g. phi-3) may have outdated remote code version kept for backward compatibility and non-maintained by its authors as the model is already integrated into transformers.

Suggestion is try to load model without remote code and only if it is failed use remote code for resolve such issues

CVS-168307

[eaidova]

@andreyanufr I do not know if this code branch can be reached at all for diffusers models, but loading config in such way does not work for diffusion models:

>>> from transformers import AutoConfig
>>> config = AutoConfig.from_pretrained("stable-diffusion-v1-5/stable-diffusion-v1-5", trust_remote_code=True)

transformers/models/auto/configuration_auto.py", line 1151, in from_pretrained
    raise ValueError(
ValueError: Unrecognized model in stable-diffusion-v1-5/stable-diffusion-v1-5. Should have a `model_type` key in its config.json

That is why I removed this line.