Add roofline data for Apple M1 (MacBook Pro 17,1)

[diegobauavi]

Summary

• Adds bench/results/rooflines/roofline-apple-m1-MacBookPro17_1-dcf5b8a7dd09-b1ab0125.json, produced by bench/roofline_suite.py --perf on an Apple M1 MacBook Pro (MacBookPro17,1), AC power.
• Regenerates bench/results/ROOFLINES.md via bench/aggregate_rooflines.py to include this machine.

Contributes to the crowdsourced ANE roofline effort in #137.

Test plan

• bench/roofline_suite.py --perf completed successfully (numeric cliffs + fast headline perf rooflines: saturation sweep, bandwidth roofline, decode measurement, roofline analysis)
• bench/aggregate_rooflines.py regenerated ROOFLINES.md without errors

[sbryngelson]

Thank you, @diegobauavi -- first fully-external roofline datapoint, and a valuable one: your M1 confirms the slice-x16 crop-DMA clamp at 4094 is present on M1 (matching M2 Pro, and contrasting M5 where it is exact), which brackets that cliff across the pre-A16 -> A16 boundary.

Your M1 data is now on main in ROOFLINES.md, credited to you on the row and as a co-author. I incorporated it via #150 rather than merging this PR directly, because after the M2 Pro submission (#145) landed, both PRs regenerate ROOFLINES.md and so conflicted on that generated file -- a mechanical collision, not anything wrong with your submission. #150 also carries a fix (_perf_headline fallback) that recovers your headline GEMM/bandwidth/ridge numbers: your run predates #147, so its roofline_analysis step errored and those columns would otherwise have been blank.

Closing as incorporated in #150. Thanks again -- please keep them coming, and sorry for the extra step on your end.