test: simplify return value of test_run_

As v0.99.5~24^2~4 (Trapping exit in tests, using return for errors,
2005-08-10) explains, callers to test_run_ (such as test_expect_code)
used to check the result from eval and the return value separately so
tests that fail early could be distinguished from tests that completed
normally with successful (nonzero) status.  Eventually tests that
succeed with nonzero status were phased out (see v1.7.4-rc0~65^2~19,
2010-10-03 and especially v1.5.5-rc0~271, 2008-02-01) but the weird
two-return-value calling convention lives on.

Let's get rid of it.  The new rule: test_run_ succeeds (returns 0)
if and only if the test succeeded.

Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Acked-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

t/test-lib.sh

@@ -457,7 +457,7 @@ test_run_ () {
 	if test "$verbose" = "t" && test -n "$HARNESS_ACTIVE"; then
 		echo ""
 	fi
-	return 0
+	return "$eval_ret"
 }
 
 test_skip () {
@@ -502,8 +502,7 @@ test_expect_failure () {
 	if ! test_skip "$@"
 	then
 		say >&3 "checking known breakage: $2"
-		test_run_ "$2" expecting_failure
-		if [ "$?" = 0 -a "$eval_ret" = 0 ]
+		if test_run_ "$2" expecting_failure
 		then
 			test_known_broken_ok_ "$1"
 		else
@@ -521,8 +520,7 @@ test_expect_success () {
 	if ! test_skip "$@"
 	then
 		say >&3 "expecting success: $2"
-		test_run_ "$2"
-		if [ "$?" = 0 -a "$eval_ret" = 0 ]
+		if test_run_ "$2"
 		then
 			test_ok_ "$1"
 		else