Enforcing a 1 to 1 matching in names between Ludwig datasets and AutoGluon paper

[abidwael]

This PR

• Modifies some of the YAMLs used to load benchmarking datasets from the AutoGluon paper. For the datasets where they submit to an online competition, they have a competition split (example). However, the scores reported in the paper are on the test split. We change the Ludwig test split for those datasets to the right file used to report scores in the paper.
• Adds the dataset news_channel (channel in the paper).
• Adds a few YAML files that allow us to use the same names in their work.

[github-actions]

Unit Test Results

         6 files  ±0           6 suites  ±0   3h 8m 20s ⏱️ - 15m 8s
  3 467 tests ±0    3 345 ✔️ ±0    77 💤 ±0  45 ❌ ±0 
10 401 runs  ±0  10 102 ✔️ ±0  254 💤 ±0  45 ❌ ±0 

For more details on these failures, see this check.

Results for commit 52255ee. ± Comparison against base commit c2fb702.

♻️ This comment has been updated with latest results.

[dantreiman]

LGTM. Also, thanks for checking the test splits against the ones used in the paper, good catch there!

[dantreiman]

I think this is the same as news_popularity - please double-check and remove the duplicate if so.

I agree with your choice to make naming match the paper, so probably news_popularity.yaml should be replaced by this one news_channel.yaml

[abidwael]

Sorry, merged before seeing this comment. Here's another PR