Now there is a request-from-bytes benchmark that supersedes
request-from-path.
Final comparison of gains made since first benchmark:
request-from-path
time: [540.01 ns 541.51 ns 543.15 ns]
change: [-90.786% -90.741% -90.696%] (p = 0.00 < 0.01)
Performance has improved.
Regression of about 100%, not clear why, but absolute change
is small. Runtime compared to bendy version is -86.407% anyway.
So save this as baseline without investigating the reason for
the regression.