Pods never go Ready on praxis 0.7.4+: probes still hit the admin port #17

Praxis 0.7.4 moved /healthy, /ready and /metrics off admin.address onto a new admin.metrics_address listener (praxis-proxy/praxis#1342). The admin port now answers those paths with 404.

We still generate admin.address: 0.0.0.0:9901 with allow_public_admin: true (src/config/generate.rs) and point the startup, liveness and readiness probes at 9901 (src/resources/deployment.rs). Any PRAXIS_IMAGE at 0.7.4 or newer gives pods that never pass the startup probe. The default image is still 0.5.2, which is the only reason nothing has broken yet.

Fix: generate admin.metrics_address: 0.0.0.0:9902, keep admin.address on loopback and drop allow_public_admin, move the probes and the container port to 9902 (the image already exposes it), and bump the default image in the same change. praxis-proxy/praxis#1285 tracks the drain and grace changes the pods will want after that.