Document the development workflow in CONTRIBUTING.md

[maragubot]

What was found

CONTRIBUTING.md explained the social process for contributing — submit a PR
directly for small changes, open an issue first for larger ones, plus the
licensing terms — but never said how to build, test, or lint the project.

The Makefile has test, lint, cover, benchmark, and fuzz targets, and
AGENTS.md documents make test, make lint, and the coverage rule. But
AGENTS.md is addressed to AI assistants, so a human contributor had no
documented way to verify a change before opening a PR. The file has had no
build or test guidance in its entire history.

What was fixed

A ## Development section covering the five make targets, the coverage
expectation, and the Go version floor.

Each claim was checked by running it:

• make test passes, and every package outside internal/ reports 100.0%.
  The section says "outside internal/" because go tool cover -func reports
  97.5% overall — internal/assert has no tests and is excluded via
  codecov.yml, which the docs did not mention anywhere.
• .golangci.yml declares version: "2", and golangci-lint v1 refuses to run
  against it, so the section asks for v2 specifically.
• make fuzz passes -fuzztime 10s per target, matching a timed run.
• The CI test matrix covers Go 1.18 through 1.26. A go 1.18 directive stops
  newer language features at compile time but silently allows newer standard
  library packages, so a green local run can still fail CI. Rather than tell
  contributors to avoid newer APIs, the section points at the convention the
  repo already follows: gomponents_benchmark_test.go and
  html/benchmark_test.go use testing.B.Loop behind //go:build go1.24.

Review notes

Two reviewers checked this by execution over two rounds. Everything they raised
is resolved. Their first pass caught the bare "100% coverage" claim
contradicting what make cover actually shows, the missing golangci-lint v2
requirement, and a tautological over-long line; the second pass caught that an
earlier draft telling contributors to "avoid newer language and library
features" contradicted the repo's own build-tagged benchmarks.

Two findings were left alone as out of scope for a single-issue PR, noted here
in case they are worth separate passes:

• AGENTS.md now duplicates make test, make lint, and the coverage rule,
  and is the less precise of the two on coverage. Worth deciding whether it
  should defer to CONTRIBUTING.md.
• internal/examples/app is a separate module with its own go.mod and
  go.work, so neither make test nor CI touches it. A contributor editing
  the example app gets no verification at all.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (76f04c0) to head (5f2c594).

Additional details and impacted files

@@            Coverage Diff            @@
##              main      #340   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            6         6           
  Lines          639       639           
=========================================
  Hits           639       639           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.