refactor: optimize OpenSearch query building in merge suggestions [CM-765]

[skwowet]

Changes

Before:

• Query building created individual clauses for each identity (60 orgs, 75 members)
• Hit OpenSearch maxClauseCount limit (1024) with organizations having 70+ identities
• Sequential query construction without deduplication
• No chunking strategy for large identity sets

After:

• Increased identity processing limit (100 for both orgs and members)
• Categorize identities into query types before building OpenSearch queries (weak/fuzzy/prefix for orgs; verified/unverified exact/fuzzy for members)
• Deduplicate fuzzy and prefix values before querying
• Chunk queries into groups of 20 to stay under clause limits
• Extract common chunkArray utility for reuse
• Use filter instead of must for tenant filtering (performance optimization)
• Change error logging from info to error for better visibility

Fixes: Organizations with 70+ identities now process successfully without hitting OpenSearch clause limits while maintaining query quality.