Reframe pyspark-test-runner default test strategy around grouped runs

[vmariiechko]

Related Issue

N/A

Summary

The pyspark-test-runner SKILL and README told the agent to pass -x by default, but -x (--exitfirst) stops pytest at the first failure, so only one failure reaches the JUnit report and the digest's signature dedup has nothing to collapse. That contradicts the digest's headline feature, which the SKILL's own description sells two sentences earlier. This makes the broad grouped run (no -x) the canonical default and documents -x as an opt-in for drilling a single suspected failure.

Docs/guidance only. The wrapper already defaults -x off and its code is unchanged.

Changes

• SKILL.md: rewrote the frontmatter description, the "Default test strategy" section (two scoped strategies, grouped run as default), the canonical invoke example (dropped -x), the flags-table row, and the FAILED "Reading the result" line.
• README.md: dropped -x from the usage example and reframed the usage sentence to match.
• databricks_template_schema.json: fixed the two -x references in the post-install success_message, including the whole-suite-plus-stop-at-first smoke-check line.
• CHANGELOG.md: Changed entry under [Unreleased], noting no wrapper behavior change.

Change Area

• Docs / tests / infrastructure only

Configuration Axes Affected

• None of the above (docs, tests, infrastructure only)

Testing

• All tests pass (pytest tests/ -V) — 2415 passed, 163 skipped

Asset Changes (if applicable)

• Asset is self-contained (no references to library/helpers.tmpl or other assets)
• Asset appears in ASSETS.md catalog

Checklist

• No .tmpl files appear in generated output
• Documentation updated (if behavior changed)