feat(client): stage native multi-resolver foundation and adaptive selector

[nightowlnerd]

Summary

• Add core state machine transition contracts in slipstream-core to establish explicit connection/resolver/stream lifecycle invariants.
• Add client-side resolver telemetry and switch-reason taxonomy, including startup switch logging and per-path health signals.
• Introduce ResolverManager as the single owner of resolver set/state, with adaptive single-active selection and guarded failover.
• Harden failover to avoid false demotion by requiring multi-signal confirmation before switching on active-path loss.
• Keep recursive standby resolvers inactive during steady state to prevent parallel-path interference.
• Add dual-resolver health telemetry and apply recursive poll throttling scoped to recursive multi-resolver sets.

Why

• Reduce resolver control-flow sprawl in runtime loop and make follow-up correctness work safer.
• Improve diagnosability with explicit health/switch telemetry.
• Move to native single-active failover behavior without round-robin workarounds.
• Prevent false failovers and stabilize dual-resolver runtime behavior under real traffic.

Validation

• cargo fmt
• cargo test -p slipstream-core
• cargo test -p slipstream-client
• cargo clippy -p slipstream-client --all-targets -- -D warnings
• Repeated canary/soak runs on remote VPS environments with real traffic patterns (anonymized), verifying:
  • startup-only resolver selection in steady state
  • no unexpected path-unavailable failover churn
  • stable single-active recursive behavior with standby inactive