Refactor manual tests job: extract reset logic and separate infinite block tests

[nhorton]

Summary

This PR refactors the manual_tests job to improve maintainability and test organization by extracting common reset logic into a reusable step and moving infinite block tests into a dedicated serial step.

Key Changes

• Version bump: Updated from 1.2.1 to 1.3.0

• New reset step: Created steps/reset.md containing centralized reset instructions that other steps can call internally

  • Consolidates git reset, file cleanup, and queue clearing logic
  • Eliminates duplication across test steps
  • Provides clear documentation of reset procedures

• New infinite block tests step: Created steps/infinite_block_tests.md as a dedicated step for infinite block testing

  • Moved 4 infinite block tests (2 prompt-based, 2 command-based) from run_fire_tests
  • Tests both "should fire" (no promise) and "should NOT fire" (with promise) scenarios
  • Runs serially with resets between tests due to blocking nature

• Updated run_not_fire_tests.md:

  • Reduced from 8 to 6 tests (removed infinite block tests)
  • Added explicit sub-agent configuration requirements: model: "haiku" and max_turns: 5
  • Updated quality criteria to reference reset step instead of inline commands

• Updated run_fire_tests.md:

  • Reduced from 8 to 6 tests (removed infinite block tests)
  • Added explicit sub-agent configuration requirements: model: "haiku" and max_turns: 5
  • Updated quality criteria to reference reset step instead of inline commands
  • Clarified that infinite block tests are handled separately

• Updated job.yml:

  • Added sub-agent configuration guidance in description
  • Updated step descriptions to reflect test count changes
  • Added new infinite_block_tests step to workflow
  • Updated changelog with version 1.3.0 entry

Implementation Details

• All test steps now reference the reset step for cleanup procedures, reducing documentation duplication
• Sub-agent configuration (model: "haiku", max_turns: 5) is now explicitly documented in all test steps
• Infinite block tests are isolated in their own step to allow for proper serial execution and controlled observation of blocking behavior
• Quality criteria have been updated across all steps to be more consistent and reference the reset step
• The reset step includes detailed explanation of each command and when to use it