chore(deps): bump cert-manager, metrics-server, headlamp and vector

[BetoFandino]

Description:

Bumps four subchart dependencies, all validated on a live cluster.

Dependency	From	To	App version
cert-manager	v1.19.3	v1.21.1	v1.21.1
metrics-server	3.13.0	3.13.1	0.8.0 → 0.8.1
headlamp	0.43.0	0.44.0	0.44.0
vector	0.50.0	0.57.0	0.53.0 → 0.57.0

Required values change

Removed vector.customConfig.api.playground. Vector 0.57 dropped that option and the binary now aborts at config load:

ERROR vector::cli: Configuration error.
  error=unknown field `playground`, expected `enabled` or `address` in `api`

Because Helm deep-merges values, the key leaked into every deployment even when operators removed it from their own values — so this broke for anyone enabling vector.

Notes for operators

Vector 0.57 — template confinement in sinks. Templates in sink config fields must now start with a literal prefix. "{{ field }}" is rejected at startup; "prefix={{ field }}" is fine. This doesn't affect the chart defaults, but it does affect operators using templates in aws_s3.key_prefix, aws_cloudwatch_logs.stream_name, file.path, HTTP sink URIs, etc.

See more in vector release notes

cert-manager 1.21 — deprecations. installCRDs is deprecated in favour of crds.enabled/crds.keep; the chart fails hard if both are set to true. This chart keeps installCRDs for backwards compatibility.

See more in Cert-manager release notes

headlamp — 0.44.0
Headlamp release notes

metric-server — 3.13.1
Metric-server release notes

Testing

Deployed on an AWS EKS cluster (Kubernetes 1.35) via kustomize with a local chart copy:

• metrics-server — deployed clean, no issues.
• cert-manager — deployed clean, no issues.
• headlamp — deployed clean, no issues.
• vector — deployed clean, no issues.