Cover compiler diagnostics in the canonical conformance suite

[konojunya]

Background

The canonical conformance suite covered normalized IR but only one compiler diagnostic. Portable diagnostic consumers need stable coverage of every Stack 1.0 code emitted by compiler stages 1 through 4 before the WASM and CLI surfaces are introduced.

Implementation

• Add eleven invalid cases covering the remaining compiler diagnostic families and deterministic multi-diagnostic ordering.
• Exercise invalid UTF-8, BOM handling, CRLF byte offsets, Unicode scalar columns, syntax failures, semantic validation, layout scope, dense-node warnings, and complexity limits.
• Mark the two encoding sources as binary so Git preserves their exact bytes.
• Add a standard-library coverage validator that requires all 22 compiler-stage codes while leaving renderer-stage diagnostics out of scope.
• Run the coverage validator in CI and document the expanded suite.

Testing

• check-jsonschema --check-metaschema schemas/*.json
• check-jsonschema --schemafile schemas/normalized-ir.schema.json conformance/valid/*/expected.ir.json
• find conformance -name expected.diagnostics.json -print0 | xargs -0 check-jsonschema --schemafile schemas/diagnostic-expectations.schema.json
• python scripts/validate-compiler-diagnostics.py
• STACK_SPECIFICATION_DIR=../specification cargo +stable test --features conformance --test conformance in stack-sh/compiler
• git diff --check

Review focus / concerns

Please review the end-exclusive byte, line, and Unicode scalar ranges, especially the binary encoding fixtures and deterministic ordering in multi-diagnostic cases. Human-readable diagnostic wording remains deliberately outside the expectation documents.