[codex] add siphon operational production profile

[haasonsaas]

Summary

• make the tap server shutdown timeout configurable instead of hardcoding 10s
• add a charts/siphon/values-production.yaml profile that enables HPA by default for production installs
• extend chart/config validation to cover the new server field and production profile rendering

Why

This is a bounded first slice of #19 covering the operational improvements only. It leaves the audit integration and protobuf publisher migration untouched.

Validation

• go test ./config ./cmd/tap -count=1
• direct runtime config surface validation for config.example.yaml and Helm-rendered config.yaml
• bash scripts/assert-chart-render.sh
• helm lint charts/siphon
• go test ./... -count=1
• go test ./... -race -count=1
• go vet ./...
• git diff --check

Notes

./scripts/lint-config.sh still uses Bash 4 lowercase expansion and fails on macOS's stock Bash 3, so I ran the script's underlying validation steps directly locally. CI runs on Ubuntu and is unaffected.

[cursor]

PR Summary

Medium Risk
Operational/config changes that affect shutdown behavior and Helm autoscaling defaults; misconfiguration could impact rollout/shutdown characteristics but does not touch security-critical logic.

Overview
Makes HTTP shutdown draining configurable by introducing server.shutdown_timeout (wired from config/env/Helm, defaulting to 10s) and using it instead of a hardcoded 10s in cmd/tap/run.go.

Adds a Helm production profile via charts/siphon/values-production.yaml that enables autoscaling (HPA) by default, updates chart schema/docs/config examples accordingly, and extends tests/scripts to validate the new config field and that the production profile renders an HPA and omits fixed deployment replicas.

^{Reviewed by Cursor Bugbot for commit 5de3ee7. Bugbot is set up for automated code reviews on this repo. Configure here.}