fix: make service ready timeout configurable via CARTESI_SERVICE_READY_TIMEOUT

[jplgarcia]

Problem

DefaultServiceTimeout is hard-coded to 5 seconds. On slow or resource-constrained environments (e.g. fly.io) the inspect-server — and occasionally other services — take longer than 5 seconds to initialise, so the supervisor kills them and the node fails to start.

Measured on fly.io: inspect-server takes ~21 seconds to become ready.

Fix

Replace the const with a package-level var that reads CARTESI_SERVICE_READY_TIMEOUT from the environment at process start:

var DefaultServiceTimeout = func() time.Duration {
    if v := os.Getenv("CARTESI_SERVICE_READY_TIMEOUT"); v != "" {
        if d, err := time.ParseDuration(v); err == nil && d > 0 {
            return d
        }
    }
    return 5 * time.Second
}()

• Fully backward-compatible: the 5-second default is unchanged when the variable is not set
• Works for both ReadyTimeout and StopTimeout fallbacks in SupervisorService.Start()
• Also affects DefaultServiceTimeout usage in http.go
• All existing tests pass

Testing

Deployed on fly.io with CARTESI_SERVICE_READY_TIMEOUT=30s:

Service is ready  service=inspect-server    (after ~21s)
All services are ready  service=rollups-node

No Service timed out errors observed.