Add standard deviation for requests

[klauspost]

The adds standard deviation to detailed analysis and compares for requests and TTFB. Examples:

Analysis:

λ warp analyze -analyze.v warp-get-2020-06-24[154208]-jmuM.csv.zst
...
Operation: GET (574228). Ran 1m0s. Size: 10000 bytes. Concurrency: 100.

Requests considered: 573756:
 * Avg: 10ms, 50%: 10ms, 90%: 15ms, 99%: 24ms, Fastest: 0s, Slowest: 117ms, StdDev: 5ms
 * TTFB: Avg: 10ms, Best: 0s, 25th: 8ms, Median: 10ms, 75th: 13ms, 90th: 15ms, 99th: 23ms, Worst: 116ms StdDev: 5ms
 * First Access: Avg: 10ms, 50%: 10ms, 90%: 15ms, 99%: 20ms, Fastest: 1ms, Slowest: 25ms, StdDev: 4ms
 * First Access TTFB: Avg: 10ms, Best: 1ms, 25th: 8ms, Median: 10ms, 75th: 13ms, 90th: 15ms, 99th: 20ms, Worst: 25ms StdDev: 4ms
 * Last Access: Avg: 10ms, 50%: 10ms, 90%: 15ms, 99%: 19ms, Fastest: 1ms, Slowest: 22ms, StdDev: 4ms
 * Last Access TTFB: Avg: 10ms, Best: 1ms, 25th: 8ms, Median: 10ms, 75th: 13ms, 90th: 15ms, 99th: 19ms, Worst: 22ms StdDev: 4ms

Compare:

λ go build&&warp cmp -analyze.v warp-get-2020-06-24[154208]-jmuM.csv.zst warp-get-2020-06-24[153319]-hBwl.csv.zst
576728 operations loaded... Done!
581688 operations loaded... Done!
-------------------
Operation: PUT
* Average: -5.22% (-0.5 MiB/s) throughput, -5.22% (-48.8) obj/s
* Requests: Avg: +4.15ms (+4%), P50: +4.6619ms (+4%), P99: +4.3328ms (+2%), Best: +3.7533ms (+17%), Worst: -82.0389ms (-24%) StdDev: +1.059646ms (+6%)
* Fastest: -4.42% (-0.4 MiB/s) throughput, -4.42% (-42.9) obj/s
* 50% Median: -4.42% (-0.4 MiB/s) throughput, -4.42% (-42.9) obj/s
* Slowest: -13.69% (-1.2 MiB/s) throughput, -13.69% (-130.1) obj/s

Throughput stddev doesn't make much sense. And since requests are reflective of throughput we have what we need.

[klauspost]

Looking for volunteers to make the output into nice tables :)