It appears as though this test didn't actually test what it was supposed
to in 25505bd ("Skip pretty formatting if the first line isn't a
TAP plan"). It was supposed to test that bats-format-tap-stream would
not try to format non-TAP data -- but instead it just showed that the
preamble of test files was executed before printing the TAP header
(something that is not going to be the case with the suite refactor).

Fixes: 25505bd ("Skip pretty formatting if the first line isn't a TAP plan")
Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>

This allows us to remove most of the complicated handling between
bats-exec-test and bats-exec-suite (which had a complicated interplay
due to the fact they both supported multi-test execution).

Instead, multi-test execution is entirely left to bats-exec-suite and
bats-exec-test only handles a single test. This entirely removes the
need for the magic output handling in bats-exec-test -- and opens the
door to suite-wide parallelism in one fell swoop.

The only downside is that bats-preprocess is run multiple times on the
same source file, but a follow-up patch will improve the caching of
pre-processed files such that they will only be processed once.

Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>

Testing that parallel execution actually works is a bit difficult, so
the most trivial way of doing it is to just have a bunch of sleep tests
and make sure that they run faster than would be possible if they were
run in serial.

Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>