…n for streaming tests

- Backend (mp_runner.py): Modified search_by_dur() to collect individual query latencies and calculate p99/p95/avg percentiles for each concurrency level
- Backend (read_write_runner.py): Updated run_search() and run_search_by_sig() to extract and propagate concurrent latency data through the streaming test pipeline
- Data Model (metric.py): Added 5 new fields to store concurrent latency data per stage: st_conc_num_list_list, st_conc_qps_list_list, st_conc_latency_p99_list_list, st_conc_latency_p95_list_list, st_conc_latency_avg_list_list
- Frontend: Created new concurrent_detail.py component with QPS vs Latency visualization featuring:
  * Single Stage view: Detailed QPS vs Latency chart with metrics table for one selected stage
  * Compare Stages view: Multi-line chart comparing QPS vs Latency evolution across multiple stages
  * Support for P99, P95, and Average latency metrics
  * Graceful handling of stages with no concurrent data
- Frontend (streaming.py): Integrated concurrent performance detail section into streaming results page

This enables users to answer 'At X QPS with writes happening, what is the latency?' for streaming workloads.