This commit updates the scaffolding to have a different behavior in
cases where we *know* that the existing project (from the yaml file) is
core24-based.

In these cases, we now raise a RuntimeError outlining the name of the
command instead of raising a ClassicFallback error to let the
non-craft-application workflow take over.

With this, Snapcraft runs on core24 projects will fail until the
commands are ported over, while projects for older bases (or commands
that don't use projects, like "init") will still fallback to the
existing behavior.

These are all the tests from tests/spread/core22, copied over to
tests/spread/core24. Completely unmodified, they therefore still have
the core22 base.

This is a semi-manual update of "base" to "core24", "build-base" to
"devel" and "grade" to "devel" (where applicable).

This commit does the equivalent of pytest's "xfail(strict=True)": all
core24 tests are currently fail because no Snapcraft command has been
implemented yet, so the test checks that the tool fails with the
expected message, and then exits successfully.

The idea is that once a command is implemented the corresponding spread
tests will start to fail - we can then scrutinize each test to validate
the new behavior.

The intention here is to speed-up spread runs and make the development
loop faster. I've run the non-core24-tests in this feature branch and
can confirm that they pass (or have unrelated failures).