support turbojpeg for mmcv imread

[song-zhang]

Motivation

When decode a 3840*2160 jpeg image, Turbojpeg is more than twice as fast as cv2. I pass backend = "turbojpeg" to imread function and get an error:

Traceback (most recent call last):
File "/home/zs/test_turbojpeg.py", line 4, in
imread(img_path, backend="turbojpeg")
File "/home/zs/mmcv/mmcv/image/io.py", line 226, in imread
return imfrombytes(img_bytes, flag, channel_order, backend)
File "/home/zs/mmcv/mmcv/image/io.py", line 268, in imfrombytes
img = jpeg.decode( # type: ignore
AttributeError: 'NoneType' object has no attribute 'decode'

Since turbojpeg doesn't support APIs such as resize, etc. I can't set turbojpeg as the default backend through use_backend("turbojpeg").

So, we need an pr to support turbojpeg for decoding jpeg image.

Modification

update "imfrombytes" implementation to support image decoding with turbojpeg backend.

BC-breaking (Optional)

No

Use cases (Optional)

Checklist

Before PR:

• I have read and followed the workflow indicated in the CONTRIBUTING.md to create this PR.
• Pre-commit or linting tools indicated in CONTRIBUTING.md are used to fix the potential lint issues.
• Bug fixes are covered by unit tests, the case that causes the bug should be added in the unit tests.
• New functionalities are covered by complete unit tests. If not, please add more unit test to ensure the correctness.
• The documentation has been modified accordingly, including docstring or example tutorials.

After PR:

• If the modification has potential influence on downstream or other related projects, this PR should be tested with some of those projects, like MMDet or MMCls.
• CLA has been signed and all committers have signed the CLA in this PR.