Evaluate CI shape changes for faster builds

[acgetchell]

Summary

Consider larger CI-shape changes that could reduce pull-request wall-clock time without reducing coverage, correctness, or platform confidence.

This is intentionally separate from the low-risk CI speedup issue for example execution, Clippy target shaping, and nextest adoption.

Candidate changes

• Split full correctness validation from portability smoke checks.
  • Keep the project-supported development platforms meaningful.
  • Evaluate whether every heavy repository-wide check needs to run identically on every OS, or whether some OS jobs should run targeted Rust portability validation plus selected full checks.
• Move benchmark compile validation out of the default main CI path or isolate it.
  • Preserve benchmark harness coverage.
  • Consider a benchmark-compile job that is path-filtered to src/**, benches/**, Cargo.toml, and Cargo.lock.
  • Avoid losing feature-gated benchmark coverage.
• Run repository tooling checks once where portability does not matter.
  • Candidates may include Markdown, TOML, YAML/CFF, Semgrep, Python tooling, shell linting, actionlint, and citation checks, depending on this repository's tooling.
  • Keep local comprehensive validation available for developer confidence.

Constraints

• Do not reduce test, benchmark, or example compile coverage without an explicit replacement.
• Preserve platform confidence for supported development environments.
• Preserve feature-gated bench/example coverage.
• Measure before/after job timings so the trade-off is visible.

Possible acceptance criteria

• CI wall-clock time improves meaningfully on pull requests.
• Required checks still catch benchmark harness, example, doctest, lib-test, integration-test, and feature-gated compile failures.
• The workflow documents which checks are portability-sensitive and which are repository tooling checks.
• The local comprehensive validation command remains available.