Add support for custom collectors

[phenixblue]

Summary

Adds a full custom collector framework that allows short-lived Kubernetes Jobs to gather node or cluster-scoped data and expose it to Rego policies via input.cluster.collectors.

New subcommand: kvirtbp collect

Deploys collector Jobs against a live cluster and writes the results to a collector-data JSON file.

kvirtbp collect \
  --bundle ./examples/collectors/node-info \
  --bundle ./examples/collectors/hugepages \
  --save-bundle ./saved-bundles \
  --output collector-data.json

Changes to kvirtbp scan

• --collector-data (repeatable) — injects pre-collected data into input.cluster.collectors
• --collector-bundle / --collector-config — run inline collectors during scan without a separate collect step
• --policy-bundle now auto-detected from _meta.bundlePaths in the collector-data file when not explicitly provided (--no-auto-bundle to opt out)
• Multi-bundle support: --policy-bundle pointing to a directory of bundles evaluates each sub-bundle and merges findings

Collector configuration

Collectors are declared in a bundle's metadata.json under a collectors array (CollectorConfig):

• scope: once — single cluster-wide Job, output at input.cluster.collectors["name"]["_cluster"]
• scope: per-node — one Job per node, output keyed by node name
• tolerations — pod tolerations so collectors can schedule on tainted nodes (e.g. control-plane)
• privileged, hostPID, hostNetwork, env for advanced collectors

Bundle improvements

• --bundle on collect is now repeatable; configs merged across bundles
• --save-bundle persists resolved (possibly remote) bundles locally; multiple bundles saved to bundle-0/, bundle-1/, …
• bundle.SubBundles() detects whether a directory is a single bundle or a container of multiple bundles
• bundle.SaveDir() copies a resolved bundle for reuse without re-fetching

Example collectors

Bundle	Collector	Scope	Checks
examples/collectors/node-info	node-info	per-node	Collector data present; CPU architecture consistency
examples/collectors/hugepages	hugepages	per-node	Collector data present; hugepages configured on all/partial/no nodes

Both include OPA unit tests (13 tests total, all passing).

Internal packages

• internal/collector — Collector interface, JobCollector, CollectorConfig, CollectorResult/CollectorMeta (with _meta envelope), MergeAll
• internal/bundle — Resolve, SaveDir, SubBundles
• internal/cli/collect.go — new collect subcommand
• docs/collectors.md — full schema and usage guide