Provide an API for manually clearing the Region Cache in long-running Spark tasks.

[lizhenhuan]

Enhancement

Users desire to expedite task processing and thus initiate a long-running Spark process, which processes approximately one billion data points every ten minutes. Initially, within the first 24 hours, the task can complete the processing of around one billion data points within ten minutes. However, as time progresses, a significant number of Regions become invalid, leading to retries and consequently causing the task's runtime to exceed ten minutes. Therefore, there is a demand to incorporate an API that allows for the manual clearing of the Region Cache after each ten-minute cycle of task completion.