Test dependencies and ordering — conditional execution chains

[ThomasMalletCodra]

Description

Allow declaring dependencies between tests so that a test is only executed if its prerequisite tests have passed. This avoids running a cascade of integration tests when the setup/environment test has already failed, saving time and producing cleaner reports.

Current behavior

• Tests are executed in discovery order (alphabetical or as listed).
• No dependency mechanism — all tests run regardless of previous results.
• A failed setup test still triggers all downstream tests, which all fail with confusing errors.

Expected behavior

• Tests can declare dependencies via # guitest: directives:

  # guitest: show
  # guitest: depends: test_setup_environment
  def test_integration_scenario():
      ...

• If a dependency fails, dependent tests are automatically marked as "skipped (dependency failed)".
• Dependency chains are visualized in the GUI (tree or graph view).
• Circular dependencies are detected and reported as configuration errors.
• Independent tests (no dependencies) can still run in any order.

Implementation ideas

1. Extend # guitest: parser to support depends: directive.
2. Build a DAG (directed acyclic graph) of test dependencies.
3. Topological sort for execution order.
4. Skip dependents immediately when a prerequisite fails.
5. GUI: show dependency arrows or indent dependent tests under their parent.

Acceptance criteria

• Tests can declare dependencies on other tests
• Dependent tests are skipped when prerequisites fail
• Skipped tests show reason: "dependency X failed"
• Circular dependencies are detected and reported
• Execution order respects the dependency graph
• Independent tests are unaffected