[Feature] Support video input and output in restoration demo

[ckkelvinchan]

Motivation

When inferencing a video model, one would need to convert the video (e.g., .mp4) to images, and then load the images. This causes inconvenience if one wants to quickly try the result.

Modification

This PR supports using video format as input and output. One could simply replace the directory with the path to the video.

Note: When saving the outputs as a video, it introduces video compression, which would lower the visual quality.

[codecov]

Codecov Report

Merging #622 (b67cf98) into master (8dce073) will decrease coverage by 0.11%.
The diff coverage is 35.00%.

@@            Coverage Diff             @@
##           master     #622      +/-   ##
==========================================
- Coverage   80.69%   80.57%   -0.12%     
==========================================
  Files         199      199              
  Lines       11383    11409      +26     
  Branches     1822     1830       +8     
==========================================
+ Hits         9185     9193       +8     
- Misses       1915     1932      +17     
- Partials      283      284       +1     

Flag	Coverage Δ
unittests	80.55% <35.00%> (-0.11%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
mmedit/apis/restoration_video_inference.py	17.85% <16.12%> (+1.19%)	⬆️
mmedit/datasets/pipelines/random_degradations.py	94.78% <100.00%> (+0.10%)	⬆️
...ls/components/stylegan2/generator_discriminator.py	84.93% <0.00%> (-1.21%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 09104fd...b67cf98. Read the comment docs.

[Yshuo-Li]

Several lines are not covered by tests

[ckkelvinchan]

Several lines are not covered by tests

It is tested in GPU only. Therefore it is not revealed in the CI report.