fix: remove convolution_2d from evidence table, clarify docstring

feat: add real rocprof CSV evidence and sync all benchmark docs (MI300X gfx942, ROCm 7.0, May 8 2026)

docs fix

feat: real MI300X benchmark results - 2.61x matmul speedup, 4077 GB/s bandwidth

docs: add live MI300X benchmark results - all 4 kernels compiled on gfx942