flan/HANDOFF-x86-cost.md

Handoff — what the x86 backend costs, and keeping it from rotting

Branch: dev-loop. This lane is item 7 of HANDOFF-x86-rt.md's "What remains" — code size and speed, measured rather than guessed — plus the closing note under it, which asks for the survey to run on its own so a refusal cannot sit unnoticed for a month again.

Nothing in lib/ is touched. Another lane is rewriting lib/x86.ml at the same time, so everything here lives in spike/x86 and in test/dune, and anything this lane finds that would want a compiler change is written down at the bottom rather than made.

The plan, in the order it is being done

1. test/dune grows an @x86 alias beside @sanitize and @valgrind, running spike/x86/survey.sh in a strict mode that exits nonzero when anything DIFFERs or is refused by name. Done first because it is the half that stops the rot and it is bounded.
2. spike/x86/cost.sh measures binary size, the program's own text, and best-of-N wall time for each corpus program both ways, and spike/x86/COST.md reads the numbers out.

Status

Stub. Nothing measured yet.