Fix benchmark noise status reporting

[justrach]

Summary

• make benchmark markdown status match the actual CI gate: percentage threshold plus absolute ns threshold
• add an absolute delta column so tiny-tool percentage swings are easier to read
• label high-percent, low-absolute rows as NOISE instead of FAIL
• add Python unit coverage for the status/report rendering

Validation

• python3 -m unittest scripts/test_compare_bench.py
• python3 -m py_compile scripts/compare-bench.py scripts/test_compare_bench.py
• manual sample: codedb_read +22.54% / +11,399 ns now renders as NOISE and exits 0

[github-actions]

Benchmark Regression Report

Thresholds: 10.00% and 50,000 ns absolute delta

NOISE means the percentage threshold was exceeded, but the absolute delta was too small to fail CI.

Tool	Base (ns)	Head (ns)	Delta	Abs Delta (ns)	Status
codedb_bundle	306489	308775	+0.75%	+2286	OK
codedb_changes	34183	30193	-11.67%	-3990	OK
codedb_deps	5139	5314	+3.41%	+175	OK
codedb_edit	4200	4766	+13.48%	+566	NOISE
codedb_find	42827	41409	-3.31%	-1418	OK
codedb_hot	57812	56980	-1.44%	-832	OK
codedb_outline	155173	158092	+1.88%	+2919	OK
codedb_read	53074	60066	+13.17%	+6992	NOISE
codedb_search	146206	144245	-1.34%	-1961	OK
codedb_snapshot	2149126	2142620	-0.30%	-6506	OK
codedb_status	242009	237537	-1.85%	-4472	OK
codedb_symbol	43152	36751	-14.83%	-6401	OK
codedb_tree	41413	40454	-2.32%	-959	OK
codedb_word	42898	38859	-9.42%	-4039	OK