feat(health): dependency-aware readiness probe with bounded checks

[yunus-dev-codecrafter]

Closes #464

Summary

This PR introduces a dependency-aware readiness probe that ensures pods only receive traffic when all required dependencies are healthy. It also separates readiness from liveness checks, following Kubernetes best practices.

Changes

• Added a dependency-aware /api/health/ready endpoint.
• Kept /api/health/live as a simple process health check.
• Added a Horizon/Soroban circuit breaker readiness probe.
• Added per-dependency latency (latencyMs) reporting for improved observability.
• Added bounded timeouts (5s) for all health checks to prevent hanging probes.
• Executed all dependency checks in parallel using Promise.all.
• Simplified health aggregation logic to automatically support new probes.
• Improved testability by making the Horizon probe fully injectable.

Benefits

• Prevents unhealthy pods from receiving traffic.
• Supports safer rolling deployments and zero-downtime releases.
• Improves observability with per-dependency latency metrics.
• Ensures health checks complete within a predictable time.
• Makes the health check system easier to extend and maintain.

Testing

• Verified readiness returns 503 when a required dependency is unavailable.
• Verified liveness remains unaffected by dependency failures.
• Verified timeout behavior for unresponsive dependencies.
• Verified circuit breaker states are reported correctly.
• Verified latency metrics are included in readiness responses.

[Baskarayelu]

dependency-aware readiness probe with bounded checks - good ops hardening so a pod with a dead dependency stops taking traffic. merging 👍