Add CodeRabbit configuration to scope automated PR reviews

[karthikvetrivel]

Description

I added a .coderabbit.yaml so that CodeRabbit reviews this repo usefully.

Two things about this repo would make a default install noisy.

• Roughly half of all pull requests are machine generated by Dependabot, Renovate, and the cherry-pick workflow.
• Vendored and generated files also make up about 90% of tracked files.

The config addresses both:

• It uses the quiet profile, so only critical and major findings are posted inline.
• It skips pull requests authored by dependabot[bot] and github-actions[bot]. A backport only replays commits that were already reviewed on main.
• It filters vendor/, generated deepcopy and clientset code, the CRD copies that make sync-crds produces, and the intentionally malformed fixtures under testdata/.
• It disables golangci-lint, which CI already runs with our own config, and yamllint, which cannot parse our Helm and Go templates.

Checklist

• No secrets, sensitive information, or unrelated changes
• Lint checks passing (make lint)
• Generated assets in-sync (make validate-generated-assets)
• Go mod artifacts in-sync (make validate-modules)
• Test cases are added for new code paths

Testing

This change adds no Go code, so there are no new code paths to cover and no Go artifacts to regenerate.

[rajathagasthya]

Added some suggestions.

[rajathagasthya]

Suggested change

	tone_instructions: "Be direct, technical and brief. No praise or emoji. Skip style and formatting, which CI already gates. Without evidence in the diff, ask a question rather than assert a defect."
	tone_instructions: "Be direct, technical and brief. No praise or emoji. Skip style and formatting, which CI already gates. Without evidence in the diff, ask a question rather than assert a defect. Name the concrete failure each finding prevents."

[rajathagasthya]

Suggested change

	Flag assertions that would still pass if the behaviour under test were
	broken, and new operand or CRD behaviour that ships with no regression
	test. Do not ask for tests covering generated or vendored code.
	Flag assertions that would still pass if the behavior under test were
	broken, and new operand or CRD behaviour that ships with no regression
	test. Flag tests that mirror the implementation instead of checking
	behavior — the expected value is computed the same way as the code
	under test, or the test only checks that a mock was called. These break
	on refactors without catching bugs.
	Do not ask for tests covering generated or vendored code.

[rajathagasthya]

Suggested change

	ClusterPolicy, GPUCluster and NVIDIADriver are released CRDs that users
	have already applied. Treat as major any removed or renamed field,
	narrowed +kubebuilder:validation, changed +kubebuilder:default, or
	changed Go type: each breaks existing custom resources on upgrade.
	New fields must be optional and carry omitempty.
	Editing these types requires regenerating the deepcopy and CRD assets,
	so say so if the PR does not also update them.
	ClusterPolicy, GPUCluster and NVIDIADriver are released CRDs that users
	have already applied. Treat as major any removed or renamed field or
	json tag, a field made required, narrowed +kubebuilder:validation,
	changed +kubebuilder:default, or changed Go type: each breaks existing
	custom resources on upgrade.
	New fields must be optional and carry omitempty.
	Every new field is permanent API surface that must be supported across
	upgrades. Ask for justification when a new field duplicates an existing
	knob, could be derived from existing fields, or configures an
	implementation detail that could live in a ConfigMap or annotation
	instead. Spec holds user intent; status holds observed state; reject
	fields that blur that line.
	Editing these types requires regenerating the deepcopy and CRD assets,
	so say so if the PR does not also update them.