Add CodSpeed continuous benchmarking

[codspeed]

Summary

Adds continuous performance measurement to MkDocs NG with CodSpeed. Every pull request now runs a benchmark suite in CI and reports how the change affects build and rendering performance.

What was added

Benchmark suite (benchmarks/) — 23 benchmarks built on pytest-codspeed, run in CPU simulation mode. The corpus is a generated MkDocs project (31 Markdown pages across 5 nested sections, plus static assets) produced from a fixed pseudo-random seed, so the workload is identical on every run and results only move when MkDocs itself changes.

File	Covers
test_build.py	Full mkdocs build (with and without use_directory_urls), Markdown population of all pages, theme template rendering
test_structure.py	get_files, get_navigation, File creation, file lookups, page rendering (regular and large page), get_toc
test_config.py	load_config with full validation, YAML parsing, loader construction
test_search.py	Search index entry creation, JSON generation, the HTML content parser
test_utils.py	Front matter parsing, title extraction, relative/normalized URLs, nest_paths, tag stripping
corpus.py / conftest.py	Deterministic corpus generation and shared fixtures

CI workflow (.github/workflows/benchmarks.yml) — runs on pull requests, pushes to main and workflow_dispatch (used by CodSpeed for backtests). It uses OpenID Connect for authentication, so no token secret is needed. Actions are pinned by commit SHA to match the convention used by the existing workflows.

Developer ergonomics — a bench Hatch environment (hatch run bench:run) so the benchmarks can be run locally like the other checks, a note in CONTRIBUTING.md, and the CodSpeed badge in README.md.

Verification

The suite was run locally under CodSpeed's CPU simulation instrument: 23 benchmarks passed and reported, including the full-site build (~1.5 s simulated), page population (~0.9 s) and theme rendering (~0.3 s). Benchmarks also pass as plain tests (pytest benchmarks, ~3 s) so they stay cheap to run outside CI. ruff check, ruff format, isort, codespell and markdownlint were run on the touched files.

Next steps

• Merge this PR so the first run on main establishes the performance baseline; subsequent PRs will be compared against it.
• Once a baseline exists, consider enabling performance checks to fail CI on significant regressions.
• Extend the suite when new performance-sensitive code paths land (for example plugin event dispatch or mkdocs serve incremental rebuilds).

[codspeed]

Congrats! CodSpeed is installed 🎉

🆕 23 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

• test_build_site: 5.2 s
• test_build_site_no_directory_urls: 5.2 s
• test_build_theme_pages: 1.1 s
• test_populate_pages: 3 s
• test_get_yaml_loader: 406.1 µs
• test_load_config: 47.3 ms
• test_yaml_load: 14 ms
• test_content_parser: 552.6 ms
• test_search_index_add_entries: 718.6 ms
• test_search_index_generate: 13.8 ms
• test_file_creation: 47.9 ms
• test_files_lookup: 3.5 ms
• test_get_files: 9.4 ms
• test_get_navigation: 11.4 ms
• test_get_toc: 416.4 µs
• test_page_render: 80.5 ms
• test_page_render_large: 447.6 ms
• test_get_markdown_title: 456.5 µs
• test_get_relative_url: 9.9 ms
• test_meta_get_data: 9.3 ms
• ...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

---