Statistical trace analytics

This task involves processing a sample of traces from production traffic and building performance model out of it. It should allow:
- detecting blocking tasks
- identifying bottlenecks and problems e.g. through critical path analysis
- building interactive performance model that would allow what-if analysis and focus on optimization of specific type of work e.g. what do I need to improve in order to improve 99th percentile latency

No files, tests, or entry points are identified in the issue, so start by locating the trace-processing and analytics entry points in the repository. Done would require a production-trace analysis capability that detects blocking tasks, identifies bottlenecks through critical-path analysis, and supports interactive what-if performance modeling.

Written by the indexing model from the issue text.