Refresh ScanCode corpus

[github-actions]

Update the embedded license corpus and conformance baseline from the latest aboutcode-org/scancode-toolkit develop branch.

Corpus regeneration report

Corpus

Metric	Previous	Regenerated	Change
ScanCode commit	6ba59089789600479cda46c84c6d436774179092	058f4396269476bdf2abf4a6b679a76f501b7e00	—
Corpus SHA-256	4a929bc0b03214a22096b9327b276ef3e7f096fa8f269dfe4e60fb35791bb1d2	d8cea48581178c38d48dacf32ba4ef642d5ac5070ffa2339b6d305e588af54eb	—
Rules	39,215	39,223	+8
Embedded size	11.55 MiB	12.17 MiB	+639.90 KiB (+5.41%)

Deterministic rebuild: pass — independently regenerated corpus contents are identical.

Cold startup

A cold New includes corpus decompression and matcher construction.

Metric	Previous	Regenerated	Change
Time	632.26 ms	649.31 ms	+2.70%
Allocated bytes	158.82 MiB	161.68 MiB	+1.80%
Retained heap	109.55 MiB	109.55 MiB	+0.00%
Allocations	526,061	543,601	+3.33%

Repeated New

Repeated New reuses the process-wide matcher engine.

Metric	Previous	Regenerated	Change
Time	36.95 ns	36.84 ns	-0.30%
Allocated bytes	17 B	17 B	+0.00%
Allocations	2	2	+0.00%

Warm Match

Warm matching reuses one matcher and runs every non-empty corpus rule as input.

Metric	Previous	Regenerated	Change
Time per corpus pass	723.83 ms	705.52 ms	-2.53%
Throughput	53.23 MB/s	54.61 MB/s	+2.59%
Allocated bytes	8.83 MiB	8.84 MiB	+0.02%
Allocations	169,728	169,760	+0.02%

Conformance

Metric	Previous	Regenerated	Change
Passed	1,578 / 1,786 (88.35%)	1,579 / 1,787 (88.36%)	+1
Divergences	208	208	0
Skipped	0	0	0

• Resolved divergences: 0
• New divergences: 0

Benchmarks are medians of five samples on Go 1.26.7 (linux/amd64). Cold startup uses one iteration per sample; repeated New and warm Match use one second per sample.