no_init_weights in from_pretrained should be optional

[IrisRainbowNeko]

What does this PR do?

After diffusers>=0.33, no_init_weights is force applied in from_pretrained to speed up loading large models. But if user add new parameters to pretrained models (like Adapters), new parameters will not be initialized and may have NaNs to crash the model.

This PR:

• Make no_init_weights in from_pretrained optional.
• Add no_init_weights and init_weights option to from_single_file to align with from_pretrained.

Crash example:

down_blocks.0.attentions.0.transformer_blocks.0.conv1.conv_out.weight torch.Size([320, 160]) tensor(nan, grad_fn=<MeanBackward0>) tensor(nan, grad_fn=<StdBackward0>)
down_blocks.0.attentions.0.transformer_blocks.0.conv1.conv_out.bias torch.Size([320]) tensor(-3.5686e+27, grad_fn=<MeanBackward0>) tensor(4.5068e+28, grad_fn=<StdBackward0>)
down_blocks.0.attentions.0.transformer_blocks.0.conv1.conv_pool.weight torch.Size([320, 320]) tensor(-5.5766e+24, grad_fn=<MeanBackward0>) tensor(1.7845e+27, grad_fn=<StdBackward0>)
down_blocks.0.attentions.0.transformer_blocks.0.conv1.conv_pool.bias torch.Size([320]) tensor(-1.7843e+27, grad_fn=<MeanBackward0>) tensor(3.1918e+28, grad_fn=<StdBackward0>)

Use example:

from torch import nn
from diffusers import UNet2DConditionModel

class MyUNet2DConditionModel(UNet2DConditionModel):
    def __init__(...):
        self.adapter = nn.Linear(...)
        ...

model = MyUNet2DConditionModel.from_pretrained(..., init_weights=True)

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline?
• Did you read our philosophy doc (important for complex PRs)?
• Was this discussed/approved via a GitHub issue or the forum? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.