Play Framework

docs(claude): fold first-run lessons into /renovate's skel gate

Three refinements to step 12, all learned running the command against
PRs 25-29:

Control before blaming the bump. On a skel-gate failure, rebuild the
scratch copy with packageManager pinned back to the previous version and
repeat. A green control names the mechanism; a failing control means the
defect predates the bump. The pnpm 10->11 finding was only trustworthy
because that isolated the single variable.

Force a real reinstall when testing a fix. pnpm answers "Already up to
date" and skips re-resolution on a tree that still has node_modules/, so
a working pnpm-workspace.yaml change looks broken -- hit mid-diagnosis,
and briefly concluded a correct fix did not work. Same false-green shape
as the vendored-jar trap the command exists to prevent.

Note that bare `pnpm` is deliberate, not an oversight. Suspected it would
run the ambient pnpm and skip the version under test; it does not --
manage-package-manager-versions defaults on, so pnpm reads the skel's
packageManager field and self-switches. Recorded so the next reader does
not "fix" it by adding a corepack prefix that changes nothing.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>