In HybridCollapsingTopDocs collector, a priority queue is created for each group and subquery combination: https://github.com/opensearch-project/neural-search/blob/main/src/main/java/org/opensearch/neuralsearch/search/collector/HybridCollapsingTopDocsCollector.java#L489

The size of each queue is set to the size parameter provided in the query, which is the same as when using hybrid query without collapse. However, we are storing a lot more documents than in hybrid search without collapse because there are multiple queues instead of a single queue.

We should add a setting that allows users to control the number of documents that are stored in each queue. Lowering this setting would prioritize latency over recall, while raising the setting would do the opposite. This setting would allow users the ability to customize for their individual use case.