What changes were proposed in this pull request?

This PR implements one pipeline write for mergetree:

1. Since ColumnBatch is returned in pipeline write, we now pass string instead of Map from C++ to java to avoid complex type, and we can utilize spark's utility to parse partition which is more compatibility.
2. Using BasicWriteJobStatsTracker to create `BasicWriteTaskStats. which compelete previous work.
3. Fix header mismatch issue: we need to get header from pipeline builder instead of query plan, since query plan only contains read rel output header.
4. [GLUTEN-7028][CH][Part-5] Refactor: add NativeOutputWriter to unify CHDatasourceJniWrapper #7395 introduced message write, we now use it in the pipeline mode for both mergetree and files(parqet, orc).
5. Introducing RutimeSettings and RuntimeConfig
6. Writing mergetree in one pipeline.

More details on RutimeSettings and RuntimeConfig

To avoid harded code in product code, We now start to define type safed ConfigEntry[T] for configuration and setting key, For example:

      .setCHConfig("path", UTSystemParameters.diskOutputDataPath)
      .setCHConfig("tmp_path", s"/tmp/libch/$SPARK_DIR_NAME")

will be changed to:

      .set(RuntimeConfig.PATH.key, UTSystemParameters.diskOutputDataPath)
      .set(RuntimeConfig.TMP_PATH.key, s"/tmp/libch/$SPARK_DIR_NAME")

(Fixes: #7028)

How was this patch tested?

(Please explain how this patch was tested. E.g. unit tests, integration tests, manual tests)

(If this patch involves UI changes, please attach a screenshot; otherwise, remove this)