Fix TWAS benchmark parsing gaps in BenchmarkMetric.js

Fix TWAS benchmark parsing gaps in BenchmarkMetric.js (SOABench key, DayTrader30/WebSphereFootprint memory regex, CPU Util pct whitespace)

## Problem
After standing up Perf_Pipeline_TWAS(https://hyc-runtimes-jenkins.swg-devops.com/job/Perf_Pipeline_TWAS/1/) , the Traffic Light table only showed 2 of 6 expected benchmark rows (WebSphereStartup + DayTrader30 Throughput). The other 4 were silently dropped -- not a missing-data issue, a parsing issue in `TestResultSummaryService/parsers/BenchmarkMetric.js`.

## Fix
Updated `BenchmarkMetric.js`:
- Added a `SOABench` key (in addition to existing `SOABENCH`) matching the exact benchmark name casing.
- Loosened CPU Util pct regex whitespace matching to non-greedy.
- Updated Adjusted Single Server Memory regex + units field from KB to MB for `DayTrader30` and `WebSphereFootprint`.

Start in TestResultSummaryService/parsers/BenchmarkMetric.js and inspect the existing benchmark keys and metric regular expressions. Verify the SOABench casing, CPU Util pct whitespace, and adjusted single-server memory handling for DayTrader30 and WebSphereFootprint. Done means the TWAS results produce all six expected benchmark rows, including the previously dropped entries.

Written by the indexing model from the issue text.