Chart: bundled Authelia's fatal NTP startup check crash-loops on clock skew, with no docs or config passthrough

[dev-dull]

Introduced in #86 (Authelia forward-auth).

Problem

The bundled Authelia (templates/authelia/configmap.yaml) uses the default config, which includes Authelia's NTP startup check — and that check is fatal. If the Kubernetes node clocks are skewed beyond Authelia's tolerance, the pod crash-loops with:

level=error msg="Error occurred running a startup check" error="the system clock is not synchronized accurately enough with the configured NTP server" provider=ntp
level=fatal msg="One or more providers had fatal failures performing startup checks" providers="[ntp]"

Two friction points:

1. Opaque failure — it presents as a generic CrashLoopBackOff; you only find the cause in the logs. Nothing in the chart docs warns that node time sync is a hard requirement.
2. No escape hatch — the config is baked into the ConfigMap template and there's no values passthrough for extra Authelia config or env, so you can't set ntp.disable_startup_check: true (or point ntp.address at a reachable server) without forking the chart.

Context

In my case this correctly surfaced a real infra problem — my nodes had no NTP service running at all and had drifted ~3 minutes — so I fixed node time sync rather than disabling the check. But the fatal-with-no-override behavior makes a common misconfig hard to diagnose.

Suggested fix (either/both)

• Docs: note in the README that the bundled Authelia requires node clocks to be NTP-synced (its startup check is fatal), with the one-line log signature above.
• Config passthrough: expose something like auth.authelia.extraConfig (merged into configuration.yml) or auth.authelia.extraEnv, so operators can set AUTHELIA_NTP_DISABLE_STARTUP_CHECK=true / a custom NTP server when appropriate. This would also generalize to other Authelia tuning.