[FEATURE] Add poll jitter to prevent thundering herd after restarts

[gergely-szabo-sap]

Context

The CF provider does not implement poll jitter. Without jitter, resources created at similar times requeue simultaneously, causing "thundering herd" spikes — especially after provider restarts when all resources are re-enqueued at once. At scale (500+ resources), this can briefly saturate both the internal rate limiter and the CF CAPI per-user rate limit.

Scope

• Add WithPollJitterHook to the controller options in cmd/provider/main.go
• Expose --poll-jitter-percentage CLI flag (following provider-kubernetes pattern, default 10%)
• Apply jitter to all controllers (all CF controllers are pure native, so no hybrid/upjet complications)
• Document how to configure in the README.md example in the BTP
• Test the jitter configuration

Out of scope

Hint: Smaller feature wishes tend to be implemented and shipped quicker!

• Adaptive poll intervals (longer-term, requires WithPollIntervalHook)
• Per-controller jitter settings

Technical Steps

1. Add --poll-jitter-percentage flag to cmd/provider/main.go (default 10, same as provider-kubernetes)
2. Compute jitter: jitter := time.Duration(float64(*pollInterval) * float64(*pollJitterPercentage) / 100.0)
3. Apply via managed.WithPollJitterHook(jitter) in controller options
4. Reference implementation: provider-kubernetes main.go exposes --poll-jitter-percentage; provider-gcp hardcodes 5%

Workarounds & Alternatives

• Manually stagger resource creation times (impractical at scale)

Additional context

• Crossplane provides WithPollJitterHook in crossplane-runtime/pkg/reconciler/managed/reconciler.go
• Recommended jitter by landscape: Small (<100) — none, Medium (100–500) — ±10s, Large (500–2000) — ±30s, XL (2000+) — ±1m

[gergely-szabo-sap]

@enrico-kaack-comp, please review this before refinement