ci: filter unaffected suites and refresh release documentation

[MeliValesca]

Documentation-only changes previously fell through to full native testing, and every main push ran all suites. Add filtering for Markdown and documentation media on PRs and main pushes. Android-only changes run Android plus library checks; iOS-only changes run iOS plus library checks. Shared code, dependencies, CI changes and unknown paths run all suites. Manual/scheduled runs and scope failures retain full coverage. Required check names and branch protection are unchanged; unaffected device entries report no tests run.

Also correct the API examples, explain web fallbacks and native command completion, and refresh the README with 25-fps accent-sliding and trackpad GIFs plus MP4 companions for iPad and Android phone.

Validation:

• Full corepack yarn check passed, including CI-helper tests (61 pass after the concurrency regressions).
• Both Python CI-helper suites passed (10 tests total).
• Actual committed README/media diff selects no code suites; the CI edits select all suites.
• Regression coverage includes mixed docs/code changes, platform-only changes, PR and push diffs, initial pushes, dependency inputs, and scope failures.
• All 11 TSX examples parse; documentation/media links resolve; fresh gesture captures visually reviewed and four GIFs verified at 25 fps.

This PR intentionally runs all suites because it changes CI. Documentation-only PRs will skip them after this merges.

Concurrency: preserve every main push in its own group so newer pushes cannot cancel running or replace pending pipelines. Runs execute in parallel or wait for runner capacity. PR runs still replace outdated runs because scope covers the full PR diff. Added regression coverage that evaluates the workflow expressions for three successive main pushes and two updates to the same PR.