fix(graph): inconsistent schema for external modules using is_external boolean instead of dedicated label

[vitali87]

Problem

The graph schema uses two different patterns for distinguishing internal vs. external entities:

• Packages: Separate node labels (Package vs ExternalPackage)
• Modules: A boolean property is_external on the same Module label

This inconsistency was noted in the gdotv blog analysis (Part 2) as a schema design flaw.

The is_external property is defined in constants.py:191 as KEY_IS_EXTERNAL = "is_external" but is not part of the documented NODE_SCHEMAS in types_defs.py, not included in the protobuf schema, and not mentioned in the Cypher system prompts. This means LLM-generated Cypher queries cannot reliably distinguish external from internal modules.

Expected Behavior

Either:

1. Create a dedicated ExternalModule label (matching the Package/ExternalPackage pattern), or
2. Unify both patterns to use boolean flags consistently

Option 1 is preferred for consistency and because it makes Cypher queries more natural (MATCH (m:ExternalModule) vs MATCH (m:Module {is_external: true})).

Affected Components

• codebase_rag/constants.py (KEY_IS_EXTERNAL)
• codebase_rag/types_defs.py (NODE_SCHEMAS)
• codebase_rag/parsers/definition_processor.py (module node creation)
• codebase_rag/schema_builder.py (schema text generation for LLM prompts)
• codec/schema.proto (protobuf schema)