tests: improve test infrastructure coverage

[garritfra]

Overview

A review of the compiler's test infrastructure identified 9 concrete gaps. This issue tracks the overall effort; each gap has a dedicated subissue.

Subissues

• tests: add unit tests for LLVM and x86 codegen backends #184 — Add unit tests for type inference (infer.rs)
• tests: add unit tests for JS codegen backend #185 — Add unit tests for JS codegen backend
• tests: add unit tests for type inference (infer.rs) #186 — Add unit tests for LLVM and x86 codegen backends
• tests: isolate integration tests that require external tools #187 — Assert on error message content in error test cases
• tests: assert on error message content in error test cases #188 — Add unit tests for AST transformation passes
• tests: add unit tests for AST transformation passes #189 — Isolate integration tests that require external tools
• tests: add coverage for module resolution edge cases #190 — Add coverage for module resolution edge cases
• tests: add property-based tests for the lexer #191 — Add property-based tests for the lexer
• tests: add benchmark suite for parser and codegen #192 — Add benchmark suite for parser and codegen

Background

The current test suite has strong coverage of the lexer and parser, and reasonable coverage of the QBE codegen backend. However, several compiler phases and backends have little to no direct test coverage, the integration test setup has reliability and portability issues, and there are no performance benchmarks.