Consider circuit breaker for AuthZen PDP client

Context

Issue #4144 introduces an AuthZen-compatible PDP integration for evaluating scopes when scope_policy: "dynamic" is configured. If the PDP endpoint becomes unavailable, every token request will wait for the full timeout before returning 503.

Problem

Without a circuit breaker, a down PDP causes all dynamic-policy token requests to hang for the timeout duration. This degrades the node's responsiveness even though the outcome (503) is already known after the first few failures.

Proposal

Add a circuit breaker to the AuthZen HTTP client so that after N consecutive failures, subsequent requests fail immediately until a probe request succeeds. This gives bounded, fast failure instead of repeated timeout waits.

Considerations

• Configuration surface: failure threshold, recovery probe interval
• State management: tracking consecutive failures per endpoint
• Whether this should be a generic HTTP client feature or specific to the AuthZen client
• May not be needed if operators ensure PDP reliability through monitoring — assess based on real-world usage after #4144 ships

Related

• #4144 (parent: mixed OAuth2 scopes with configurable scope policy)

Start by reading issue #4144 and the AuthZen HTTP client it introduces. Assess the failure behavior and resolve the open questions around threshold, probe interval, per-endpoint state, and whether the breaker belongs in the generic client or AuthZen client; done means fast failure after repeated PDP failures and recovery after a successful probe.