Added start_test() and finish_test() functions to test/common.sh.in.
See usage information in the (new) cleanup() function in that file.
This allow us to isolate test failures based on naming of individual
tests within a single test file. As part of this, added set -e to
setup_test(). This required minor updates to test/new/basic and
test/pick/test-stderr to conform. So, migrated test/new/test-basic
to start_test()/finish_test().
(cherry picked from commit
cd6133dd23422f92534328f768a322f63a5c6dfe)