188f54ea

By: Tom Sydney Kerckhove <syd@cs-syd.eu>

Correct the measured figures and drop em dashes

The first speed numbers were single unmeasured runs and one memory
figure belonged to a different run entirely. Re-measured interleaved
with repeats: fix's bulk advantage is a range rather than a point
because it is parallel and the machine is shared, and on real flakes it
is consistently slower than what we have.

Also removes feature-progress.md, which was scaffolding.
GitHub
nix-ci-evalnix-ci-eval-integration-testnix-ci-eval-shallow-fetch-testnix-ci-eval-stack-depth-testnix-ci-evalnix-ci-eval-integration-testnix-ci-eval-shallow-fetch-testnix-ci-eval-stack-depth-testconfiguredevShelldefaultnix-ci-evaldefaultnix-ci-evalaarch64-linuxx86_64-linux