Run the Go test suites verbosely so CI shows what executed

[stevebeattie]

Why

A bare go test prints only ok <package>. That cannot distinguish a suite
that ran from one that silently covered less than expected — a newly added
table row or fixture that never runs looks exactly like one that passes.

Not hypothetical here: the offline matrix pairs fixtures with package-level
blobs, which Go does not report as unused, so a row dropped in a refactor leaves
a compiling, green, quieter suite. Nothing in the log would say so.

The immediate prompt was #164. It passes the mirror check in CI, but the log
showed only ok — so the thing that PR is about, the compared-file count
going from 7 to 8, could not be confirmed from the run at all.

Change

-v on test-offline and validate_mirrors. One-line each, plus a comment
explaining why.

What the logs now show

    --- PASS: TestOfflineFixtureMatrix/certificate_audit/fail_java_missing_stamp
    ... 47 matrix rows, each named
    mirrors_test.go:390: compared 7 standalone OVAL file(s) against the datastream
    mirrors_test.go:393: present only in the datastream, so not evaluable with
        `oscap oval eval` and with no input for regeneration: oval:org.stub:def:1

The matrix row count and the mirror check's file count are both now readable
from a CI run, so a drop in either is visible rather than inferred.

Verification

make validate_mirrors and make test-offline both pass: 169 named subtests,
47 matrix rows, 0 failures. The full offline log grows to 846 lines, which seems
a fair trade for being able to tell a green run from a quiet one.

Note

Touches the Makefile, as does #166. The edits are in different regions
(lint-go versus these two recipes) so they should merge cleanly in either
order, but worth knowing if both are in flight.

🤖 Generated with Claude Code