Cache before partitioned-write, provide unpersist handle

[EnricoMi]

This works around SPARK-40588.

[github-actions]

Test Results

   144 files  ±    0     144 suites  ±0   22m 9s ⏱️ +55s
   351 tests +  19     351 ✔️ +  19  0 💤 ±0  0 ❌ ±0 
4 574 runs  +266  4 574 ✔️ +266  0 💤 ±0  0 ❌ ±0 

Results for commit d595b78. ± Comparison against base commit 74c8075.

This pull request removes 1 and adds 20 tests. Note that renamed tests count towards both.

uk.co.gresearch.spark.SparkSuite ‑ UnpersistHandle does unpersit set DataFrame

uk.co.gresearch.spark.SparkSuite ‑ NoopUnpersistHandle does not unpersist set DataFrame with StorageLevel(1 replicas)
uk.co.gresearch.spark.SparkSuite ‑ NoopUnpersistHandle does not unpersist set DataFrame with StorageLevel(disk, 1 replicas)
uk.co.gresearch.spark.SparkSuite ‑ NoopUnpersistHandle does not unpersist set DataFrame with StorageLevel(disk, memory, deserialized, 1 replicas)
uk.co.gresearch.spark.SparkSuite ‑ NoopUnpersistHandle does not unpersist set DataFrame with StorageLevel(memory, deserialized, 1 replicas)
uk.co.gresearch.spark.SparkSuite ‑ SilentUnpersistHandle does not throw on unpersist if no DataFrame is set
uk.co.gresearch.spark.SparkSuite ‑ SilentUnpersistHandle does not throw on unpersist with blocking if no DataFrame is set
uk.co.gresearch.spark.SparkSuite ‑ SilentUnpersistHandle does unpersist set DataFrame with StorageLevel(1 replicas)
uk.co.gresearch.spark.SparkSuite ‑ SilentUnpersistHandle does unpersist set DataFrame with StorageLevel(disk, 1 replicas)
uk.co.gresearch.spark.SparkSuite ‑ SilentUnpersistHandle does unpersist set DataFrame with StorageLevel(disk, memory, deserialized, 1 replicas)
uk.co.gresearch.spark.SparkSuite ‑ SilentUnpersistHandle does unpersist set DataFrame with StorageLevel(memory, deserialized, 1 replicas)
…

♻️ This comment has been updated with latest results.