fix: apply filter expressions in --dry-run and make -qq silent

[MatthewMckee4]

Summary

karva test --dry-run -E <expr> printed every collected test regardless of the filter expression. The dry-run branch was calling collect_tests and print_collected_tests before FiltersetSet::new had even been constructed — the parse call further down was only validating user input and then throwing the result away. This change parses the filterset first, stores it, and threads it through collect_test_names so each qualified name is evaluated against an EvalContext before being printed. Tag filters still degrade to an empty tag set because tag metadata is a runtime Python concept and is not available at collection time, but the common test(...) predicates now work as documented.

-qq was silently equivalent to -q even though Verbosity exposes quiet as ArgAction::Count and VerbosityLevel::Silent is already wired through Printer::stdout_important to disable the failure-summary stream. Verbosity::level() only checked self.quiet > 0 and returned Quiet, so Silent was unreachable. The fix mirrors the existing self.verbose match:

match self.quiet {
    0 => {}
    1 => return VerbosityLevel::Quiet,
    _ => return VerbosityLevel::Silent,
}

cli_arg() already returns "-qq" for Silent, so the flag is correctly propagated to workers via inner_cli_args without further changes.

Test Plan

• dry_run_applies_filter_expression in crates/karva/tests/it/filterset.rs — asserts that --dry-run -E test(~alpha) against a two-test file prints only test::test_alpha and 1 tests collected.
• qq_is_silent_not_quiet in crates/karva/tests/it/basic.rs — asserts that -qq on a failing run produces empty stdout, where the existing test_quiet_output_failing (invoked with -q) still prints the summary line.
• just test — 826 passed, 0 failed.
• uvx prek run -a — all hooks green.

[codspeed-hq]

Merging this PR will not alter performance

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

✅ 1 untouched benchmark

---

_{Comparing fix/dry-run-filter-and-qq (8e23ba2) with main (c05b0d3)}