index planning: Refine cost models after initial phase of testing

[dimitarvdimitrov]

Validate and refine the cost models for index lookup planning. Focus on assumptions around regex matching, posting intersections, and count-min sketch accuracy/size trade-offs.

Test and analyze current cost models, enhance them where assumptions do not hold, and document findings.

more specifically

• Incorporate matcher deduplication to lookup planning #12831
• lookup planning: adjust plan's intersection cost #12830
• labels: Adjust cost estimates mimir-prometheus#990
• Use square root selectivity reduction instead of assuming independence (similar to SQL server's "exponential back-off")
• Adjust cost function for regex (.|||(.||||z||:||D){|z||||(.||||z||:||D){|:||D){974}
• More accurately estimate .*val1.*|.*val2.*|.*val3.* regex matcher expense
• index planning: add local benchmarks against TSDB blocks and "query stats" queries #13041