[Data] Modify ImageDatasource to use Image.BILINEAR as the default image resampling filter.

[ronyw7]

Why are these changes needed?

Ray Data's current ImageDatasource uses PIL's resize function as the image processing backend, which by default uses the Image.BICUBIC resampling filter. In practice, we found this is around 20% slower than using the Image.BILINEAR filter, which is the default option in torch vision.

Here are some benchmark results:

• https://gist.github.com/ronyw7/655255425359874321a1b9009b8446a1
  This is a unit test where we load one single image and resize it, repeated 10000 times. Note the time difference between the current (Image.BICUBIC) filter vs. proposed (Image.BILINEAR).

    Example Output:
    Time taken with default (BICUBIC) implementation: 45.74734696099995
    Time taken with BILINEAR mode: 40.40636501500012
    .
    ----------------------------------------------------------------------
    Ran 1 test in 86.155s
    OK

• https://gist.github.com/ronyw7/811ee71a0f8d2daa8d2945a2eec8c247
  This is an end-to-end benchmark modified from @stephanie-wang 's image loader microbenchmark. Here, we demonstrate the actual effect on the ray data image loading pipeline. Using the BILINEAR filter as default leads to a ~22% increase in throughput.

    - Default (BICUBIC)
    ray_data,1137.3931758979813
    ray_data,1146.6874856299994
    ray_data,1147.080762189496
    ray_data,1153.9882179831754
    ray_data,1140.6351740671025

    - Proposed (BILINEAR)
    ray_data,1399.06477460088
    ray_data,1381.023590432492
    ray_data,1394.170545465624
    ray_data,1396.9916050718184
    ray_data,1389.6461855214884

If a user still wishes to use the BICUBIC filter, this is still easily achievable by applying a UDF resize_fn after the images have been read. For instance, we can choose cv2's INTER_CUBIC or PIL's original resize (this requires the use of PIL's fromarray though, as the output of read_image are numpy arrays; this conversion lowers throughput).

• Demo: https://gist.github.com/ronyw7/f070c00d2e3965397c37d7b991b20143

Related issue number

Checks

• I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
  • I've added any new APIs to the API Reference. For example, if I added a
    method in Tune, I've added it in doc/source/tune/api/ under the
    corresponding .rst file.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[c21]

Thank you @ronyw7!