Add CodSpeed performance benchmarks

[codspeed-hq]

Summary

• Added performance benchmarks for the lexer, parser, and bytecode builder using google_benchmark through the CodSpeed compatibility layer
• Created a dedicated CodSpeed GitHub Actions workflow that runs benchmarks in simulation mode on every push to master and on pull requests
• Added the CodSpeed badge to the README

Changes

Benchmarks (bench/)

Three benchmark files covering the main components:

• lexer_bench.cpp: Measures lexing throughput on simple, medium, and complex programs
• parser_bench.cpp: Measures parsing throughput (with pre-lexed tokens) and combined lex+parse performance
• bytecode_builder_bench.cpp: Measures bytecode emission for arithmetic ops, jumps, and a simulated loop construct

Build system (CMakeLists.txt)

• Added ENABLE_BENCHMARKS option (default OFF) to build benchmarks independently of tests
• google_benchmark is fetched via FetchContent from CodSpeed's compatibility layer, which provides instrumented measurements in CI
• The benchmark executable pxkorka_benchmarks links against korka_lib and benchmark::benchmark

CI workflow (.github/workflows/codspeed.yml)

• Runs on ubuntu-24.04 with GCC 14 (matching the existing build workflow)
• Builds with -DCODSPEED_MODE=simulation and RelWithDebInfo for meaningful profiling data
• Uses OIDC authentication (tokenless) via CodSpeedHQ/action@v4

Next steps

• After merging, performance reports will appear on future pull requests automatically
• Visit CodSpeed to view performance trends and flame graphs
• Consider adding more benchmarks as the compiler and VM runtime components are completed

[codspeed-hq]

Congrats! CodSpeed is installed 🎉

🆕 10 new benchmarks were detected.

You will start to see performance impacts in the reports once the benchmarks are run from your default branch.

Detected benchmarks

• BM_ParseSimpleProgram: 8.6 µs
• BM_EmitJumps: 3.6 µs
• BM_LexSimpleProgram: 6 µs
• BM_ParseComplexProgram: 27.3 µs
• BM_LexAndParseComplexProgram: 45.3 µs
• BM_LexMediumProgram: 11.8 µs
• BM_ParseMediumProgram: 16.7 µs
• BM_EmitArithmeticOps: 2.8 µs
• BM_BuildComplexBytecode: 5 µs
• BM_LexComplexProgram: 26.2 µs

---