Fix bug in non-distributed multi-gpu training/testing

[MeowZheng]

Since MMDP does not support non-distributed multi-GPU training, --gpus in train.py lost its role, so it is removed, --gpu-ids is changed to --gpu-id, because only one GPU can be specified for non-distribution training and testing.
If the number of GPUs is more than 1, there will be an assertion error in MMDP because MMDP does not support the situation with more than one GPU.

ref : open-mmlab/mmdetection#7019

[codecov]

Codecov Report

Merging #85 (96ace96) into master (573d79c) will decrease coverage by 0.06%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master      #85      +/-   ##
==========================================
- Coverage   68.08%   68.02%   -0.07%     
==========================================
  Files          86       86              
  Lines        4888     4888              
  Branches      704      704              
==========================================
- Hits         3328     3325       -3     
- Misses       1453     1456       +3     
  Partials      107      107              

Flag	Coverage Δ
unittests	68.02% <ø> (-0.07%)	⬇️

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

Impacted Files	Coverage Δ
mmflow/datasets/pipelines/transforms.py	79.33% <0.00%> (-0.50%)	⬇️

---

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 573d79c...96ace96. Read the comment docs.