Adds Ray implementation of IntegratedGradientsExplainer that distributes across cluster resources

[tgaddair]

• Adds RayIntegratedGradientsExplainer to do label-parallel explanations (as datasets tend not to be the bottleneck in most cases)
• Adds support for doing global explanations, which returns a single explanation in the list: the mean explanation across all rows
• Sets max_calls=1 for Ray tasks that use GPUs due to https://docs.ray.io/en/latest/ray-core/tasks/using-ray-with-gpus.html?highlight=max_calls#workers-not-releasing-gpu-resources

[github-actions]

Unit Test Results

         6 files  ±  0           6 suites  ±0   3h 43m 34s ⏱️ + 6m 38s
  3 500 tests +  6    3 377 ✔️ +  6    79 💤 ±0  44 ❌ ±0 
10 392 runs  +15  10 117 ✔️ +15  231 💤 ±0  44 ❌ ±0 

For more details on these failures, see this check.

Results for commit 424041c. ± Comparison against base commit 467eab6.

♻️ This comment has been updated with latest results.