Simplify the logic of script sort parallel collection

[KunjueYu]

This pr introduced a simpler way to support parallel collection for script sort.
The concurrency issue of script sort parallel collection was fixed in Fix concurrency issue in ScriptSortBuilder, which introduced a concurrentMap to save the relationship between leafContext and leafScript, so that the correct leafScript can be obtained later. However, we could fix the concurrency issue in another way, which may get rid of the map and simplify the code.

Before Fix concurrency issue in ScriptSortBuilder, the leaf script was a member of anonymous subClass of BytesRefFieldComparatorSource or DoubleValuesComparatorSource, so that it's shared across the shard and is not thread-safe.
If we create a named subClass of BytesRefFieldComparatorSource or DoubleValuesComparatorSource, we can put the declaration of leaf script member in the FieldComparator (for BytesRefFieldComparatorSource) or LeafFieldComparator (for DoubleValuesComparatorSource). So that different comparator could hold a different reference of leaf script.
In this pr, the leaf Comparator isn't wrapped as Fix concurrency issue in ScriptSortBuilder did because in each slice, the leafs are searched sequentially. So that we could set leaf script as a member of the FieldComparator, and the leaf script will be assigned to a new values for each leaf context and consumed later in the setScorer method, which works the same way just as before.

[cla-checker-service]

💚 CLA has been signed