[BUG] Update spark.rapids.sql.reader.multithreaded.combine.sizeBytes recommendation to default

[parthosa]

Currently, the AutoTuner recommends 10m for spark.rapids.sql.reader.multithreaded.combine.sizeBytes as compared to the plugin default. We should update the heuristic to:

1. Introduce a rule for updating the metric
2. Disable recommendation (so that plugin default gets used)

Ref: https://nvidia.github.io/cudf-spark/docs/additional-functionality/advanced_configs.html