What does this PR do?

Fixes bug in DDIMInverseSampler
The current sampler has not aligned the timestep with prediction, this caused two problem:

1. the previous step may be underfloor, this is easy to validate, and we can print the previous step and current step.

#     1. get previous step value (=t+1)
        prev_timestep = timestep
        timestep = min(timestep - self.config.num_train_timesteps // self.num_inference_steps, self.num_train_timesteps - 1)
        print("prev_t:",prev_timestep.item(), end=" ")
        print("curr_t:",timestep.item(), end="\n")

the print results looks like:

2. the next problem is the noise is predicted at the current step, but using the alpha and beta of the previous step.
   as the above code snippet suggests, the current timestep is overwritten by the code line, thus delivers wrong result.

timestep = min(timestep - self.config.num_train_timesteps // self.num_inference_steps, self.num_train_timesteps - 1)

3. I think the previous name convention is not suitable in inverse problem, maybe next_timestep and timestep better.
4. the results visualization
   the code book link: colab
   I use pix2pixZero demo to demonstrate:
   the misaligned sampler:
   
   this pr:
   

Fixes # (issue)

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.