Improve conformance baseline reporting

[lucarlig]

What changed

• use one cargo-nextest-style colored report for local make conformance and CI
• distinguish XFAIL (expected failure reproduced), FAIL (expected pass failed), and XPASS (expected failure passed)
• classify pinned fixture findings as informational UPSTREAM results so they do not fail the dataplane job or appear in dataplane failure details
• add make conformance-bless to refresh dataplane-owned expected failures from the current run
• isolate each local run's result directory and add focused reporter regression coverage

Why

The official runner's raw status combined dataplane baseline mismatches with known upstream fixture findings, while the local target did not run the detailed reporter. That made it hard to see which expectation changed and could fail CI for an upstream issue.

The shared reporter is now authoritative in both environments. Missing results remain an error, but pinned upstream findings are reported without being counted as dataplane failures.

Usage

Run the conformance suite and compare its results with the checked-in expectations:

make conformance

After an intentional behavior change, run the suite and refresh the expected-failure baseline:

make conformance-bless

Review and commit the resulting change to tests/conformance/expected-failures.yml.

Verification

• tests/conformance/report-baseline-diff-test.sh
• shellcheck tests/conformance/report-baseline-diff.sh tests/conformance/report-baseline-diff-test.sh tests/conformance/run-conformance.sh tests/conformance/run-local.sh
• actionlint .github/workflows/mcp_conformance.yml
• pinned full make conformance: 59 passed, 44 expected failures, 0 failed, 0 unexpected passes
• make pre-commit

[cafalchio]

I am not sure id make conformance bless is working.
But make conformance is more human readable.