fix: class-generator --update-schema misses duplicate kinds across API groups

[myakove]

Problem

When multiple CRDs share the same kind (e.g., Backup exists in both velero.io and postgresql.cnpg.noobaa.io), the class-generator has no way to let the user specify which API group they want. Running class-generator -k Backup may generate the wrong one.

Current Behavior

• class-generator -k Backup --dry-run generates for an unintended API group
• class-generator -k backups.velero.io --dry-run fails because -k only accepts kind names
• No --api-group or similar option exists to disambiguate
• After --update-schema, the mapping may not even contain the kind, yet it still picks one

Expected Behavior

Users should be able to specify which API group to use when a kind exists in multiple groups.

Reported in

#2654 (comment)

Requirements

• Add support for specifying API group when kind is ambiguous (e.g., --api-group velero.io or -k Backup.velero.io)
• When multiple API groups exist for the same kind and no group is specified, prompt the user to choose
• Update help text and documentation

Deliverables

• CLI option for API group selection
• Interactive prompt when kind is ambiguous
• Tests for disambiguation logic
• Updated CLI help text

[myakove]

Root Cause Analysis

The issue is in identify_missing_resources() (class_generator/core/schema.py:162-228).

The Bug

Missing resource detection compares using KIND only, ignoring the API group:

# Line 176-191: Gets resources from cluster
kind = parts[-1]  # KIND is the last column (e.g., "Backup")
cluster_resources.add(kind)  # No group info stored

# Line 193-228: Compares against existing mappings
kind = gvk_item.get("kind")  # Just "Backup" without group
existing_kinds.add(kind.strip())

Impact

If postgresql.cnpg.noobaa.io/Backup is already in the mappings, velero.io/Backup on the cluster is incorrectly considered "already covered" and never fetched. This is a silent failure — no warning is logged.

Why Storage/Generation Is Fine

• The mapping file already stores arrays per kind (multiple schemas per kind)
• parse_explain() groups schemas by API group and returns all variants
• generator.py creates separate files with API group suffixes (e.g., dns_config_openshift_io.py)

Fix Plan

Change identify_missing_resources() to compare using (group, kind) tuples instead of just kind name. This ensures all API group variants are detected as missing and fetched.

Affected Code

• class_generator/core/schema.py - identify_missing_resources() (lines 162-228)
• Potentially find_api_paths_for_missing_resources() (lines 301-403) if it also uses kind-only matching