refactor: Consolidate cross-cutting concerns and reduce code duplication

[jade-codes]

Summary

Analysis of the syster-base crate reveals several cross-cutting concerns and duplications that should be consolidated for better maintainability.

Identified Issues

1. Duplicate Type Definitions

Position/Span duplication:

• core::span::Position and semantic::types::diagnostic::Position - SAME type defined twice
• core::span::Span and semantic::types::diagnostic::Range - SAME concept with different names
• core::span::SymbolReference and semantic::symbol_table::symbol::SymbolReference - SAME type defined twice

Recommendation: Consolidate all position/span types into core::span and re-export from semantic where needed. The Range type should be an alias or import from Span.

2. Parallel KerML/SysML Implementations

The following files have nearly identical structure with only language-specific types swapped:

SysML	KerML	Lines	Similarity
adapters/sysml/folding.rs	adapters/kerml/folding.rs	83/82	~95%
adapters/sysml/selection.rs	adapters/kerml/selection.rs	130/127	~90%
adapters/sysml/inlay_hints.rs	adapters/kerml/inlay_hints.rs	137/109	~85%
adapters/sysml/helpers.rs	adapters/kerml/helpers.rs	137/34	~70%
adapters/sysml/visitors.rs	adapters/kerml/visitors.rs	285/226	~80%
adapters/sysml/validator.rs	adapters/kerml/validator.rs	109/47	~60%
syntax/sysml/parser.rs	syntax/kerml/parser.rs	157/63	~70%
sysml_adapter.rs	kerml_adapter.rs	identical structs	~100%

Recommendation: Create a generic/trait-based approach:

• Define a LanguageAdapter<T> trait
• Create shared implementations with type parameters
• Reduce boilerplate while maintaining type safety

3. File I/O Scattered Across Modules

File loading concerns are spread across:

• core::file_io - low-level file operations
• project::file_loader - re-exports from core + adds collection
• syntax::parser - uses file_io for parsing

Recommendation: Clarify the layers:

• core::file_io - pure file operations (no language awareness)
• syntax::parser - language-aware parsing (owns parse dispatch)
• project::file_loader - higher-level workspace operations (remove re-exports)

4. Error Types Fragmentation

Error-related types are scattered:

• core::error_codes - error code constants
• core::parse_result - ParseError, ParseErrorKind
• semantic::types::error - SemanticError, SemanticErrorKind
• semantic::types::diagnostic - Diagnostic, Severity

Recommendation: Consider a unified error hierarchy in core::errors:

core::errors/
  codes.rs        # Error code constants
  parse.rs        # ParseError
  semantic.rs     # SemanticError  
  diagnostic.rs   # Diagnostic (unified format)

5. Re-export Chains

Several modules have long re-export chains that obscure the source:

• project::mod.rs re-exports from core
• semantic::mod.rs has 15+ re-exports
• This creates confusion about where types are defined

Recommendation:

• Document the canonical import path for each type
• Reduce re-exports to only the most common use cases
• Use explicit paths in internal code

Proposed Consolidation

Phase 1: Type Deduplication (Low Risk)

1. Remove semantic::types::diagnostic::Position - use core::Position
2. Alias Range to Span or unify them
3. Remove duplicate SymbolReference

Phase 2: Adapter Generalization (Medium Risk)

1. Create LanguageAdapter<T> trait
2. Create FoldingExtractor<T> trait
3. Create SelectionExtractor<T> trait
4. Implement for both SysML and KerML

Phase 3: Module Restructuring (Higher Risk)

1. Consolidate error types
2. Simplify file_loader module
3. Reduce re-export chains

Success Criteria

• No duplicate type definitions
• Parallel language implementations share common code via traits
• Clear module boundaries documented
• cargo clippy passes without dead_code warnings
• All existing tests pass

Related

• chore: Clean up KerML and SysML grammar implementations #649 - Grammar cleanup
• feat: Specification compliance audit for KerML and SysML grammars #650 - Specification compliance audit