Parse Server crash on FCM network error using http/2

New Issue Checklist

• I am not disclosing a vulnerability.
• I am not just asking a question.
• I have searched through existing issues.
• I can reproduce the issue with the latest versions of Parse Server and the Parse Server Push Adapter.

Issue Description

The http/2 implementation in firebase-admin throws on network errors instead of handling them, which lets them propagate to the node process, which eventually crashes the process.

• For GOAWAY errors a fix has been implemented with https://github.com/parse-community/parse-server-push-adapter/issues/340.
• For other network errors such as ECONNRESET there is no easy remediation.

The proper solution for any network error is to fix https://github.com/firebase/firebase-admin-node/issues/2789, for example via an event listener.

Steps to reproduce

This issue occurs on network error events, such as ECONNRESET.

Actual Outcome

Unhandled error crashes app.

Expected Outcome

Error should be handled by the firebase-admin.

Workaround

• Use the older version HTTP/1.1 by setting fcmEnableLegacyHttpTransport: true in your push options.
• Catch the error in a global Node process error handler.

Environment

Client

• Parse Server Push Adapter version: 6.8.0

Server

• Parse Server version: 7.3.0

Start with the firebase-admin HTTP/2 implementation and the network-error discussion in Firebase issue 2789, focusing on ECONNRESET events and the event-listener approach mentioned here. Reproduce the failure with the FCM transport enabled; done means network errors are handled without an unhandled error crashing the Node process.