refactor: Remove ExecutionDataCache in favor of react-query

[schottra]

This is another step toward refactoring our data fetching for the ExecutionDetails view to use react-query instead of ExecutionDataCache.
This PR removes ExecutionDataCache and updates all code using it to use queries instead. It also includes a fair amount of supporting code and refactoring to clean up our usage of react-query.

Notes

• This is being merged into a feature branch. There are currently some TODOs and failed/disabled tests which will be addressed in follow-up PRs.
• The plan for mocking data in tests/storybook is to switch to using mock-service-worker to mock network requests. The data mocking in tests has always been a little messy/cumbersome. The test philosophy for this repo has leaned more toward integration-style testing, so letting a real network request happen and returning mocked data from a local server is more in line with that philosophy.

Full change list:

• Updated to react-query v3
• Refactored code for creating react-query resources
  • Added custom hash function to normalize non-plain objects (for use with protobufJs-generated messages classes in id fields of entities)
  • Added a retry handler which will not retry 401/404 errors.
• Removed ExecutionDataCache and related context
• Created nodeExecutionQueries file with queries/hooks related to fetching lists of node executions
• Updated code for fetching node execution details to use queries
  • Details are fetched for a NodeExecution once and then cached, since they will not change
• Added queries for Workflows and Tasks
• Full workflow and task queries will be cached indefinitely because the data is immutable
• Added global observers for Workflow, NodeExecutionList and TaskExecutionChildList queries to extract nested entities and store them in the query cache. This allows subsequent queries for a single NodeExecution or TaskTemplate to hit the cache instead of requesting data we already have. This was the primary purpose of ExecutionDataCache.
• Updated ExecutionNodeViews and child components (including DetailsPanel) to be compatible with queries.
• Updated tests/stories enough to make them compile, disabled suites for any code which previously used ExecutionDataCache. They will need a more significant change to the way data is mocked and will be handled in a future PR.
• Updated row, child, and cell rendering components for NodeExecutionsTable to get their data from a query, which should be cached from fetching the list at the level above.
• Removed loading logic from NodeExecutionsTable. There is additional work to make it handle errors and loading states correctly, but it doesn't need full details of how we fetch data.
• Updated code for fetching NodeExecution children to use queries instead of ExecutionDataCache. This resulted in less-complicated code, so the queries were moved to live inside of nodeExecutionQueries
• Added QueryType enum for use as the first value in QueryKey arrays. Each type of entity we fetch or generate will have its own type, which is combined with an id and/or RequestConfig object to generate the unique hash for the query.
• Updated low-level api calls for listing NodeExecutions, TaskExecutions, and their children to have no limit. We usually always want all items with no pagination.
• General cleanup: removing unused code and imports