Lazy initialize FixedBitSet in LRUQueryCache [LUCENE-10120]

Basing on the implement of collecting docIds in DocsWithFieldSet, may be we could do similar way to cache docIdSet in **LRUQueryCache#cacheIntoBitSet(BulkScorer scorer, int maxDoc)** when docIdSet is density.

In this way , we do not always init a huge FixedBitSet which sometime is not necessary when maxDoc is large


---
Migrated from [LUCENE-10120](https://issues.apache.org/jira/browse/LUCENE-10120) by Lu Xugang (@LuXugang), updated Mar 10 2022
Attachments: [1.png](https://apache.github.io/lucene-jira-archive/attachments/LUCENE-10120/1.png), [LUCENE-10120.patch](https://apache.github.io/lucene-jira-archive/attachments/LUCENE-10120/LUCENE-10120.patch)
Pull requests: https://github.com/apache/lucene/pull/422, https://github.com/apache/lucene/pull/422, https://github.com/apache/lucene/pull/422, https://github.com/apache/lucene/pull/422, https://github.com/apache/lucene/pull/422, https://github.com/apache/lucene/pull/422, https://github.com/apache/lucene/pull/422

Start with LRUQueryCache#cacheIntoBitSet(BulkScorer scorer, int maxDoc) and compare its doc ID collection with DocsWithFieldSet. Review the attached LUCENE-10120.patch and determine whether FixedBitSet can be allocated lazily for dense docId sets; done means avoiding unnecessary large allocation while preserving cache behavior.

Written by the indexing model from the issue text.