Text Indices Are Force Merged to Single Lucene Segment

We force merge all segments of a single Lucene index to a single segment for text indices today, thus potentially creating a single large segment and restricting the possibility of searching multiple segments concurrently. Instead, we should be depending on Lucene's merging policy and define a custom policy if we need to keep open file handles under control.

No file, test, or entry point is named. Start by locating where Pinot force-merges all segments for text indices and read the Lucene merge-policy configuration; done means text indices rely on Lucene's merging policy without forcing one segment, while controlling open file handles if needed.

Written by the indexing model from the issue text.