Add reporting visualization command

[ftshijt]

Summary

• add a packaged versa-visualize command for HTML/CSV/Markdown reports from result JSONL files
• expand aggregation reporting with mean/std, confidence intervals, failure counts, rankings, outlier examples, and multiple export formats
• add a tiny reporting example and docs for a basic smoke check

What to expect

Run:

versa-visualize demo/reporting_example_results.jsonl \
  --out /tmp/versa-reporting-example/report.html \
  --csv /tmp/versa-reporting-example/report.csv \
  --markdown /tmp/versa-reporting-example/report.md \
  --group-by model

The HTML report opens with four summary counters: utterances, numeric metrics, metric categories, and missing/invalid metric values. It then shows a radar overview, category sunburst, category data-completeness table, optional per-metric ranking by group, the detailed metric summary table, and outlier examples.

For the bundled example, the category summary should show complete coverage:

Category	Metrics	Observed Values	Missing	Invalid	Coverage
asr_wer_cer	1	6	0	0	100%
pitch_f0	1	6	0	0	100%
similarity	1	6	0	0	100%
speech_enhancement	2	12	0	0	100%

CSV/Markdown exports contain one row per metric, with columns for count, missing, invalid, mean, median, std, stderr, ci95_low, ci95_high, min, max, higher_is_better, best_key, best_value, worst_key, worst_value, and outliers.

Example CSV rows:

metric	category	count	mean	std	95% CI	best	worst
mcd	pitch_f0	6	5.65	0.9975	4.852 to 6.448	utt_005 (4.5)	utt_003 (7.1)
pesq	speech_enhancement	6	2.633	0.6022	2.151 to 3.115	utt_005 (3.4)	utt_003 (1.9)
wer	asr_wer_cer	6	0.14	0.05933	0.09253 to 0.1875	utt_005 (0.07)	utt_003 (0.23)

Tests

• .codex-test-venv/bin/pytest test/test_reporting.py
• .codex-test-venv/bin/python versa/bin/visualize.py demo/reporting_example_results.jsonl --out /private/tmp/versa-reporting-example/report.html --csv /private/tmp/versa-reporting-example/report.csv --markdown /private/tmp/versa-reporting-example/report.md --group-by model
• python3 -m py_compile versa/__init__.py versa/reporting.py versa/bin/visualize.py versa/bin/aggregate_results.py
• git diff --check