Test the JUnit dialects against their schemas, their own output and each other

[eda-s-claude-bot]

New Features

• tests/unit/JUnitDialects/ is a new test level for behaviour that only exists between a writer and a reader:
  • Schemas.py - every reference output in tests/data/JUnit validates against its dialect's schema, and the
    dialect's reader accepts what its schema accepts.
  • RoundTrip.py - per dialect: read a reference report, write it in the same dialect, validate the result and
    read it back, comparing test case count, test case names and hostnames.
  • Translation.py - the 5x5 conversion matrix: every dialect read and written as every dialect, each cell
    validated against the target schema and read back with the target reader.
• tests/README.md writes down the testing strategy: the two phases (instantiation, combination), the four
  structural levels (package / module / class / method) and how they map onto this repository, why tests/data is
  evidence rather than fixtures, and the known-gap convention below.

Changes

• Level (c) is one class per dialect, derived from a base class holding the checks; the base sets
  _dialectName = None and skips itself, so each check is written once and reported per dialect.

• Format limits are asserted, not skipped. Translation.py::FORMAT_LIMITS names each pair with the data the
  target format needs and the source lacks, and the test asserts the conversion still fails:

  ("Any-JUnit", "pyTest-JUnit"):
      "pyTest-JUnit requires 'timestamp' on <testsuite>; the OSVVM report has none.",

  A skip goes quiet forever; this way lifting a limit turns the expectation red and the entry is removed by the
  pull-request that lifts it. A conversion failing for any other reason is a defect and does not belong in that
  table.

Known Issues

• The six conversions in FORMAT_LIMITS fail correctly but not cleanly: Any-JUnit → GoogleTest-JUnit
  raises AttributeError: 'NoneType' object has no attribute 'isoformat', and pyTest-JUnit → CTest-JUnit
  writes a document that its own schema rejects rather than refusing. Fixed separately - a writer that cannot
  express a report should say so, not crash or emit invalid XML.
• TestsuiteSummary.Aggregate() does not compute a start time from its children, which is why pyTest-JUnit →
  CTest-JUnit is a limit at all: pytest's <testsuites> has no timestamp while its <testsuite> children do.
  Deriving it would turn two of the six limits into working conversions.

Unit Tests

tests/unit: 98 → 155 passed, 65 subtests, 11 skips (the base classes). The 5x5 matrix is 25 testcases.

---

Related Issues and Pull-Requests

• Follows #162. Building this matrix is what found the last
  defect fixed there: Ant-JUnit4 and CTest-JUnit dropped the hostname of their root <testsuite> element.
• The schema corrections these tests make safe - hostname as use="required" where every reference output has
  it, and Any-JUnit accepting a <testsuite> root - are deliberately left for a follow-up.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 31.72%. Comparing base (be86c31) to head (af97ec0).
⚠️ Report is 3 commits behind head on dev.

Additional details and impacted files

@@             Coverage Diff             @@
##              dev     #163       +/-   ##
===========================================
- Coverage   72.12%   31.72%   -40.41%     
===========================================
  Files           9        9               
  Lines        2490     2490               
  Branches      395      395               
===========================================
- Hits         1796      790     -1006     
- Misses        516     1700     +1184     
+ Partials      178        0      -178     

Flag	Coverage Δ
unittests	31.72% <ø> (-40.41%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

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