refactor(core): model executor spawn failures explicitly

[errfld]

What changed

• replaced the ambiguous NonSuccessfulExitStatus error with CommandSpawnFailed { command, source }
• added command redaction for spawn failures (<args redacted>) to improve diagnostics without leaking sensitive args
• refactored executor internals to use an internal command runner abstraction so spawn and exit failure paths are explicitly modeled
• kept non-zero exit and signal-termination failures distinct (ExecutionFailed, ExecutionTerminatedBySignal)
• replaced flaky shell/timing-based executor tests with deterministic fake-runner tests that cover:
  • spawn failures
  • non-zero exits
  • signal termination
  • sequential-before-parallel behavior
  • failure propagation that skips parallel hooks after sequential failure

Validation

• cargo test --workspace

Closes git-smee-7ud.

Summary by CodeRabbit

• Refactor

  • Reworked hook/command execution into a modular, pluggable runner approach to improve extensibility and testability.

• Bug Fixes

  • Improved error reporting for command spawn failures, including redaction of sensitive tokens and clearer underlying error propagation.
  • Preserved existing exit-code semantics while hardening edge-case handling.

• Tests

  • Expanded coverage for execution ordering, spawn errors, redaction, and mixed parallel/sequential hooks.

[coderabbitai]

📝 Walkthrough

Walkthrough

Introduced a pluggable CommandRunner abstraction and PlatformCommandRunner adapter, refactored command execution to use runners, added command redaction and a CommandSpawnFailed error variant, and updated tests to exercise runner-based execution and redaction behavior.

Changes

Cohort / File(s)	Summary
Executor core crates/git-smee-core/src/executor.rs	Added CommandRunner trait and PlatformCommandRunner implementation; refactored execution flows to execute_hook_with_runner, run_hooks_with_runner, and generic execute_command using a runner; implemented redact_command and is_inline_env_assignment; replaced previous error variants with CommandSpawnFailed { command, source }.
Tests / Fake runner crates/git-smee-core/src/executor.rs (tests section)	Added FakeRunner for deterministic testing; new tests for command redaction, spawn errors, empty commands, and parallel vs sequential hook execution ordering.

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~45 minutes

Poem

🐰🥕 I hopped through runners, light and fleet,
Commands now masked so secrets sleep sweet.
Platform or fake, I bounce on the cue,
Hooks run in order — then hop — and whoo! 🎩

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 47.62% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'refactor(core): model executor spawn failures explicitly' directly and clearly summarizes the main change: replacing ambiguous error handling with explicit spawn failure modeling through the CommandSpawnFailed variant.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch git-smee-7ud/executor-error-modeling

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@crates/git-smee-core/src/executor.rs`:
- Around line 114-125: The redact_command function currently treats the first
whitespace token as the executable and can leak inline env assignments like
KEY=VALUE; update redact_command to skip any leading tokens that match an
ENV_ASSIGN pattern (tokens containing '=' before any whitespace and not starting
with '-' or containing a path separator) and consider the first non-ENV_ASSIGN
token as the executable (or use "<redacted>" if none remain) before applying the
existing truncation and "<args redacted>" logic; reference the redact_command
function and the local variables tokens, executable, and redacted when making
the change.

[errfld]

Addressed the CodeRabbit comment about potential secret leakage in redaction.

• Updated redact_command to skip leading inline env-assignment tokens (KEY=VALUE) when selecting the executable token.
• Falls back to <redacted> if no executable remains.
• Added regression test: given_spawn_error_with_env_prefix_when_executing_then_redaction_hides_env_assignments.

This is now in commit 828a6cd.