fix(engine): coerce Python literal "True"/"False"/"None" in workflow output

[PolyphonyRequiem]

Problem

Workflow output templates pass through _maybe_parse_json (engine/workflow.py:3398) to convert JSON-shaped strings back into native Python types. The function only recognized lowercase JSON literals (true / false / null).

In practice, workflow author expressions like:

output:
  matched: "{{ left == right }}"
  done:    "{{ count >= threshold }}"

render their bool results via Jinja's default str(), producing the strings "True" / "False". These then survived _maybe_parse_json unchanged — as truthy non-empty strings — so downstream route when: clauses comparing them against true / false silently misbehaved. The string "False" is truthy and == true is also false, so behavior was hard to reason about and harder to spot.

A {{ none }} expression has the same shape: renders as "None", survives as a string.

Fix

Recognize the three Python-literal forms ("True" / "False" / "None") explicitly before the existing JSON-literal check. Lowercase JSON literals continue to coerce as before.

Three lines of behavior change in _maybe_parse_json:

if stripped == "True":
    return True
if stripped == "False":
    return False
if stripped == "None":
    return None

Tests

Added 3 cases to TestWorkflowEngineOutputTemplates:

• test_output_template_python_bool_literals — verifies {{ a == b }} / {{ a != b }} produce native bool.
• test_output_template_python_none_literal — verifies {{ none }} produces native None.
• test_output_template_lowercase_json_literals_still_work — regression check for true / false / null.

All 77 tests in test_engine/test_workflow.py pass; ruff clean.

Why this matters

The mismatch between Jinja's str(bool) output and the JSON-literal recognition list was a stable footgun: workflows lint clean, validate clean, run without errors, and silently take wrong route branches. Even authors who knew about it had to reach for awkward workarounds like {{ (a == b) | string | lower }} or {{ 1 if a == b else 0 }}.

This patch is intentionally small and additive — no behavior change for any input that already coerced correctly. Happy to adjust scope if you'd prefer a different approach (e.g. coerce at template-render time rather than at output-map time).

[PolyphonyRequiem]

The previous CI run surfaced 2 failing tests. Both were asserting the broken behavior this PR fixes (with inline comments acknowledging the workaround):

• test_examples.py:214 — result["syntax_passed"] == "True" # Templates return strings
• test_parallel_workflows.py:410 — result["success"] == "True" # Boolean rendered as string

Pushed 4a61195 updating both to assert native bool (is True).

This is also a useful signal: those tests effectively codified the bug as expected behavior. Worth flagging to anyone reviewing — the change is intentional, not an oversight in the assertions.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (main@45f682d). Learn more about missing BASE report.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #139   +/-   ##
=======================================
  Coverage        ?   84.68%           
=======================================
  Files           ?       53           
  Lines           ?     7255           
  Branches        ?        0           
=======================================
  Hits            ?     6144           
  Misses          ?     1111           
  Partials        ?        0           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[jrob5756]

LGTM. Surgical fix with clear docstring and good test coverage (new behavior + None + regression for lowercase JSON literals). The two updated existing tests honestly correct previously-asserted-but-buggy behavior.