Fix failing interactive exec test

Due to a change of the typical-project fixture, a test of the interactive
exec feature started failing. Arguably, this test was bad to begin with,
relying on the frailty of the typical-project fixture to simulate a failing
build. Instead, we simply change some of the code in the typical-project
fixture in the test, to provoka a build failure and thus create the context
for the interactive exec test.

test/cli/expected/interactive-exec--_ez-build_--interactive__echo_success____should_only_execute_command_on_successful_builds--1

@@ -3,9 +3,11 @@ js – src/a.js -> lib/a.js,lib/a.js.map
 success!
 js – src/added.js -> lib/added.js,lib/added.js.map
 success!
-js – src/b.js: Unexpected token, expected { (8:7)
+js – src/added.js -> lib/added.js,lib/added.js.map
+success!
+js – src/b.js: Unexpected token, expected ; (8:18)
   6 | }
-  7 |
-> 8 | export all from foo
-    |        ^
-  9 | export default 'b'
+  7 | 
+> 8 | export default 'b'export all from foo
+    |                   ^
+  9 | 

test/cli/interactive-exec.bats

@@ -24,6 +24,7 @@ teardown() {
     rm *.{pid,log}
   fi
 
+  git checkout .
   unload_fixture typical-project
 }
 
@@ -45,6 +46,6 @@ teardown() {
 }
 
 @test "'ez-build --interactive \"echo success!\"' should only execute command on successful builds" {
-  touch src/b.js
+  echo "export all from foo" >> src/b.js
   eventually 'assert_expected "$(cat build.log)"'
 }