fix: HPA metrics + PDB selector, and add missing config options

[davidusken]

Fixed bugs

1. HPA emits no metrics. The {{- if }} guards in hpa.yaml still use the old flat names (autoscaling.targetCPUUtilizationPercentage). 412be8b moved the body to the nested target: shape and ba811e1 moved values.yaml, but the guards were missed — they evaluate nil, so metrics: renders empty.

Defaults happen to work (the HPA API falls back to 80% CPU), but any custom CPU target and any memory target is silently ignored. Present since 4.2.0.

2. PDB matches zero pods. pdb.yaml uses kellnr.labels for its selector, which includes helm.sh/chart and app.kubernetes.io/version — labels the pods don't carry. Switched to kellnr.selectorLabels, now matching the Deployment.

Missing config
The chart exposed 67 of kellnr's 84 settings. Added the rest: proxy.{url,index,api,connect/requestTimeoutSeconds}, the whole gcs backend (absent despite S3 being supported), and oauth2.{enforced,autoRedirect}. Also fixed the proxy.numThreads comment (said 20, source default is 10) and the singular toolchain_bucket default.