Keep Linux and Windows workflows alive and fail regeneration on drift

[MisterGC]

The Linux and Windows workflows each get a keepalive job (liskin/gh-workflow-keepalive@v1) that runs only on their weekly schedule and re-enables its own workflow. The (sources regeneration) check step now also regenerates into src and fails when that changes the committed sources.

• keepalive job with actions: write in build_linux.yml and build_windows.yml; the action keeps only the workflow it runs in alive, so each scheduled workflow carries its own
• after zserio tutorial.zs -python build/gen, the step runs zserio tutorial.zs -python src -ignoreTimestamps + git add --intent-to-add --all -- src + git diff --exit-code -- src; new files count as drift too
• -ignoreTimestamps is needed: without it zserio skips output files that are newer than tutorial.zs, so a hand edit passes the check

Verified:

• "regeneration fails on a non-empty git diff": in a fresh clone with pip zserio==2.19.0, the committed sources → exit 0; a committed body edit in src/tutorial/role.py → exit 1 with the flag, exit 0 without it (skipped 6 files); an uncommitted new file → exit 1
• PR CI on 60bba3a: 10/10 build jobs pass (build_linux 3.9–3.13, build_windows 3.8–3.12), including the new diff check; both keepalive jobs skipped on pull_request as intended
• actionlint on both workflows → exit 1, 6 findings, all on the existing checkout@v3, setup-java@v3, setup-python@v4 lines, none on lines added here
• 2 files changed, 22 insertions(+)

Not verified: the keepalive job itself. It runs only on schedule, so its first real run is the next Sunday cron.

Closes #2