Inconsistent input and disparity in Kitti2015Stereo (and others)

🐛 Describe the bug

In the official archive for Kitti2015Stereo, there are twice as input images for each camera than there are disparities.
Matching inputs and disparities now seems to depend solely on sorting the file names in respective directories which is not enough and inconsistent input image and disparities are served as a result.
The dataset also reports size 400 but iteration ends after 200 examples due to missing disparities.

A workaround is to get rid of the extra images in extracted archive. This would be good to add into docs.
A better solution would be to use the file names to get matching counterparts, e.g., to use the file names intersection from the relevant directories as examples to serve.

It affects any stereo dataset if sorting the files is not enough.

How to check/reproduce:

import matplotlib.pyplot as plt
import torch
import torchvision
import torchvision.transforms.functional as F

# Using https://s3.eu-central-1.amazonaws.com/avg-kitti/data_scene_flow.zip
dataset = torchvision.datasets.Kitti2015Stereo('data')
print(dataset)

for i, (left, right, disparity, _) in enumerate(dataset):
    print(f'{i} / {len(dataset)}')
    left = F.to_tensor(left).mean(dim=0, keepdim=True)
    right = F.to_tensor(right).mean(dim=0, keepdim=True)
    disparity = torch.as_tensor(disparity)
    grid = torchvision.utils.make_grid([left, right, disparity])
    plt.imshow(grid.numpy().transpose((1, 2, 0)))

Versions

Collecting environment information...
PyTorch version: 1.13.0+cu117
Is debug build: False
CUDA used to build PyTorch: 11.7
ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.1 LTS (x86_64)
GCC version: (Ubuntu 11.3.0-1ubuntu1~22.04) 11.3.0
Clang version: Could not collect
CMake version: version 3.22.1
Libc version: glibc-2.35

Python version: 3.10.6 (main, Nov 14 2022, 16:10:14) [GCC 11.3.0] (64-bit runtime)
Python platform: Linux-5.15.0-58-generic-x86_64-with-glibc2.35
Is CUDA available: False
CUDA runtime version: No CUDA
CUDA_MODULE_LOADING set to: N/A
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

Versions of relevant libraries:
[pip3] numpy==1.23.4
[pip3] torch==1.13.0
[pip3] torchvision==0.14.0
[conda] Could not collect

cc @pmeier