Improve coverage reporting: separate unit vs integration test runs

[johnmccrae]

Context

Write-CoverageSummary.ps1 (ex10) runs all tests including integration tests to collect coverage. Unit tests use heavy mocking so they contribute 0 executed commands, making the reported percentage misleading. Integration tests are the only meaningful signal for coverage but are slow (~50s). A separate lightweight pass over unit tests would give a faster feedback loop.

Acceptance Criteria

• Write-CoverageSummary.ps1 accepts a -TestType parameter: Unit, Integration, or All
• Default is All to preserve existing behaviour
• Markdown summary labels which test set was used for coverage
• Both modes verified locally and documented in script comment block

Relevant Code Paths

• scripts/Write-CoverageSummary.ps1 — main script to update
• tests/Unit/ — unit tests (mock-heavy, low coverage signal)
• tests/Integration/ — integration tests (real Plaster invocations, high coverage signal)