core: handle target crash at any point

[connorjclark]

#11840 added a specific error code when the chrome target crashes, but it only applied to navigations within the waitFor condition check. Any other runner, or during teardown in navigation, and a crash would still result in a PROTOCOL_TIMEOUT.

This PR adds the crash promise to ProtocolSession, and checks during any sendCommand that there was not a crash.

Alternatively, we can introduce a wrapper around gatherFn and check if the Driver is in a crash state. This PR has both approaches (thus Draft).

[paulirish]

bringing fatal promise over to the session is awesome. works great. let's go with that

for my own posterity.. i test this with adding this to gatherers/trace.js right after the Tracing.end line:

      setTimeout(() => {
        session.sendCommand('Page.navigate', {url: 'chrome://crash'});
      }, 100);

[connorjclark]

bringing fatal promise over to the session is awesome. works great. let's go with that

OK

I realized the Driver shouldn't be the thing handling this crash promise. I just moved it to ProtocolSession, dropped all the wiring being used to get it there, and added a session.onCrashPromise() for use by navigation's wait-for-conditions.

[connorjclark]

Oh, this new approach also means we handle crashes in all sessions, not just the root.

[paulirish]

not gonna lie this is like.. sooooooooooooooo good.

net negative line count change.. and expanded the functionality. and its easier to follow.

just a huge win all around.

tested it a few ways and it looks good. i get an LHR regardless, whenever the crash happens.

NIIIIIIIIIIIIIIIIIIICE