ci: correctly clean up server process after e2e tests (#1666)

* use TERM signal to kill server on exit

using KILL does not actually kill all child processes of `yarn start`

* fail build on e2e failure

Composer/scripts/e2e.sh

@@ -6,12 +6,14 @@ yarn start &
 SERVER_PID=$!
 
 npx cypress run --browser chrome
+EXIT_CODE=$?
 cleanup
 
 function cleanup {
-  kill -9 $SERVER_PID
+  kill $SERVER_PID
 }
 
 # kill server process
 trap cleanup EXIT
 
+exit $EXIT_CODE

azure-pipelines.yml

@@ -45,7 +45,6 @@ jobs:
       - script: ./scripts/e2e.sh
         displayName: Run E2E Tests
         workingDirectory: Composer
-        continueOnError: true
         env:
           # CYPRESS_RECORD_KEY: $(CYPRESS_RECORD_KEY)
           CYPRESS_VIDEO: true
@@ -55,7 +54,7 @@ jobs:
           TERM: xterm
           COMPOSER_BOTS_FOLDER: $(Pipeline.Workspace)/Composer/TestBots
       - task: PublishPipelineArtifact@1
-        condition: in(variables['Agent.JobStatus'], 'SucceededWithIssues', 'Failed')
+        condition: failed()
         continueOnError: true
         inputs:
           targetPath: $(Build.ArtifactStagingDirectory)/cypress