Upgrade ray tune component

[ruizhuanguw]

This PR adds the following features:

• Ray tune distributed component.
  • Add the basic version.
• Ray tune component.
  • The function apply_sweep_settings had a typo for the truncation termination policy. Fix the bug.
  • Use distributed component to set up ray cluster with more than 1 node.
  • Support quniform, randint and qrandint for ray tune search space distribution.
  • Use fixed_params for lgbm training parameters instead of config. Fix the bug.
  • Misc changes: (1) update default value for max_concurrent_runs. (2) remove additional non-lgbm parameters for lgbm parameters parser. (3) add header argument in the spec yaml file.
  • Support PopulationBasedTraining, HyperBandForBOHB, TuneBOHB, OptunaSearch for ray tune component.
  • Allow low cost conifg for blend search in ray tune component.
  • Log best trial metrics during tuning process.
  • Allow ray tune logs shown in output files of AzureML workspace.
• Lightgbm training
  • Add pipeline tags for sweep, ray tune and ray tune distributed.

Addendum: this had to upgrade dependencies in the docs build to pass

[github-actions[bot]]

Unit Test Results for Build

122 tests  +5   122 :heavy_check_mark: +5   1m 52s :stopwatch: +40s     1 suites ±0       0 :zzz: ±0      1 files   ±0       0 :x: ±0 

Results for commit c3a8d0f6. ± Comparison against base commit 6fa7c5d8.

:recycle: This comment has been updated with latest results.

[github-actions[bot]]

Package	Line Rate	Branch Rate	Complexity
.	100%	0%	0
common	81%	0%	0
pipelines.azureml	68%	0%	0
scripts	100%	0%	0
scripts.data_processing.generate_data	92%	0%	0
scripts.data_processing.lightgbm_data2bin	95%	0%	0
scripts.data_processing.partition_data	92%	0%	0
scripts.inferencing.custom_win_cli	94%	0%	0
scripts.inferencing.lightgbm_c_api	75%	0%	0
scripts.inferencing.lightgbm_python	95%	0%	0
scripts.inferencing.lightgbm_ray	95%	0%	0
scripts.inferencing.treelite_python	94%	0%	0
scripts.model_transformation.treelite_compile	92%	0%	0
scripts.sample	93%	0%	0
scripts.training.lightgbm_python	76%	0%	0
scripts.training.lightgbm_ray	88%	0%	0
scripts.training.ray_tune	82%	0%	0
Summary	82% (2165 / 2646)	0% (0 / 0)	0