[srpatcha]

ebuild test reports success for a project with no tests.

result = subprocess.run(argv, cwd=str(cwd))
if result.returncode != 0:
    raise SystemExit(result.returncode)
log.success("All tests passed.")

ctest exits 0 when it finds nothing to run. A CMakeLists.txt with
enable_testing() and no add_test() still produces a
CTestTestfile.cmake, so _resolve_test_runner() finds ctest, ctest prints
No tests were found!!!, exits 0, and the wrapper calls it a pass.

Measured on current master, against a project with a configured build tree and
zero registered tests:

$ ctest --test-dir _build
No tests were found!!!
ctest exit=0

$ ebuild test
[ok] All tests passed.
exit=0

That is worse than having no test command, because it actively reassures. Anyone
wiring this into CI gets a green gate over a repository that has stopped testing
itself, and nothing anywhere says so.

The change

An empty run is a failure. Two independent checks, because neither is reliable
alone: the marker phrase catches ctest, which prints no summary at all in that
case, and a zero total catches a runner that prints a well-formed summary adding
up to nothing.

Output is now captured rather than inherited — the exit status alone cannot
distinguish "every test passed" from "there were no tests" — and echoed, so the
terminal reads as it did before.

Counts are reported when the runner printed a summary this parser recognises and
withheld when it did not. A number derived from a zero exit status is a guess
presented as a measurement. make has no standard summary format, so its counts
stay unknown and the exit status carries the verdict.

Verified

All three outcomes, on real CMake projects:

zero tests      [error] ctest completed without running a single test.   exit=1
1 pass 1 fail   [error] Tests failed (ctest exited 8).                   exit=8
1 pass 0 fail   [ok] All tests passed (1 passed, 0 failed).              exit=0

tests   287 -> 307 passed, 1 skipped, 0 failed

Also here: the docs must describe commands that exist

A check that every ebuild command shown in the docs is one the CLI registers.
Reads only fenced code blocks — prose says things like "ebuild is a unified build
system", and a plain grep over the README yields ebuild and, ebuild can,
ebuild is.

Verified it fails rather than merely existing: putting ebuild verify in the
README turns it red.

FAILED test_a_documented_command_exists[verify-README.md]

It also asserts the eight §7.1 golden-path steps directly, so the path cannot be
quietly broken by editing the README instead of the CLI.

Supersedes most of #84

#84 added test and monitor when neither existed. Both have since landed on
master from another PR, so that work is redundant — this keeps only the parts
that are not: the empty-run guard, the count reporting, and the docs check. I
will close #84 once this is reviewed.