Add LCS load generator implementation

[Balatripura587]

Description

HTTP load generator for LightSpeed Core Service (LCS) using Locust. Runs query (POST /v1/query) and streaming (POST /v1/streaming_query) endpoints sequentially in a single invocation, collecting per-endpoint performance metrics.

Run modes:

• CLI — run directly from source with ./lcs-load-generator run --host ... --users ... --duration ...
• Container — all configuration via environment variables, ENTRYPOINT runs the test automatically
• OpenShift Job — single manifest (config/lcs-load-generator.yaml) with RBAC + Job, deploy with oc apply -f

Metrics collected per endpoint:

• Latency percentiles (p50, p95, p99), min/max/avg
• Throughput (req/s), total request count
• HTTP status code distribution
• bytesIn/bytesOut
• TTFT (time to first token) and stream duration percentiles (streaming only)

Prometheus metrics scraping

Uses kube-burner (pinned to v1.10.4) to query Prometheus and index infrastructure metrics. Two metric profiles:

• metrics-report.yaml — instant aggregates (avg/max CPU, memory, throttling)
• metrics-timeseries.yaml — time-series data (per-pod CPU/memory over time, LCS application metrics, node network, histograms)

All metric profiles use {{.LCS_NAMESPACE}} template variable — works with any namespace, no hardcoding required.

Note: LCS does not have an operator, so Prometheus monitoring requires manual setup depending on where LCS is deployed.

• If deployed in an openshift-* namespace, platform Prometheus (prometheus-k8s) scrapes metrics directly
• If deployed in a user namespace, user-workload monitoring must be enabled and metrics are queried via Thanos Querier (PROMETHEUS_BACKEND=thanos)

Result indexing:

• When ES_SERVER is configured, results are indexed to Elasticsearch
• Otherwise, results are written as JSON files under RESULTS_DIR (defaults to /results in the pod, /tmp locally) and printed to stdout (visible via oc logs on OpenShift)

Additional features:

• Multi-process mode (--processes) at high concurrency
• Cluster metadata collection (OCP version, platform, worker node count) for jobSummary document
• -D (debug) and -W (quiet) logging flags for troubleshooting

Testing

1. Tested in CLI , container and deployed as OpenShift job: