Add merge script to vegeta module

[meecethereese]

This PR adds the functionality of merging results following a distributed load test setup.

• Updated vegeta run pipeline (modules/vegeta/run/run.sh)
  • Replaced temp-file-then-replay approach with a streaming tee pipeline
  • Now tee simultaneously saves raw binary to a .bin file and streams through vegeta encode
• Added multi-pod merge script (modules/vegeta/merge/merge.sh)
  - New script that takes one or more raw vegeta .bin files and produces jaggr-format per-second JSON output
  - Uses gawk with actual request timestamps for bucketing (not wall-clock time)
  - Works correctly on saved/replayed data and correctly interleaves results from pods that started at slightly different times
• Wired merge into Dockerfile entrypoint and added gawk to the base image
• Expanded vegeta module tests (modules/vegeta/test/test.sh)
  - Added tests for .bin file production, per-second bucketing correctness, single-file merge, simultaneous multi-file merge, and synthetic latency percentile verification with exact expected
  values
• Added CI multi-pod merge job (.github/workflows/validate.yaml)
  - New test-merge job that runs 4 simultaneous vegeta attacks at 50 RPS for 15s, merges the results, and validates line count, combined RPS, code histogram integrity, and JSON structure
  - Added gawk to CI module-test dependencies
  - Server readiness loop fails explicitly with container logs on timeout

[OliverMKing]

Consider adding set -o pipefail here (i.e. set -eo pipefail).

The merge pipeline is vegeta encode | sort | gawk. Without pipefail, if vegeta encode fails on a corrupted .bin file, sort and gawk will process empty/partial data and exit 0, producing incorrect (or empty) output silently. The subsequent empty-file check would only catch the fully-empty case, not partial/corrupt output.

Same applies to run.sh (pre-existing).

[meecethereese]

This is a really good catch, I'll update for this