[IA-2355] support maxRetry

[Qi77Qi]

support optional subscriber name and optional deadLetterPolicy

Tested in console:

qi | INFO: Subscriber Successfully received data: "\"ooo\""
attributes {
  key: "googclient_deliveryattempt"
  value: "5"
}
message_id: "501898938879014"
publish_time {
  seconds: 1605568147
  nanos: 653000000
}
.

After 5 retries, msg is forwarded to dead letter topic that I configed, I see the nacked message as expected.

PR checklist

• For each library you've modified here, decide whether it requires a major, minor, or no version bump. (Click here for guidance)

If you're doing a major or minor version bump to any libraries:

• Bump the version in project/Settings.scala createVersion()
• Update CHANGELOG.md for those libraries
• I promise I used @deprecated instead of deleting code where possible

In all cases:

• Replace the appropriate version hashes in README.md and the CHANGELOG.md for any libs you changed with TRAVIS-REPLACE-ME to auto-update the version string
• Get two thumbsworth of PR review
• Verify all tests go green (CI and coverage tests)
• Squash commits and merge to develop
• Delete branch after merge

[rtitle]

Just curious - since these are optional could it be a non-breaking change if they default to None?

[Qi77Qi]

ah yea, I guess (I'm not entirely sure by providing default is backward compatible since I vaguely remember I had to fix things after a PR that added optional param with default...)...I tend to not giving default just so caller have to think about these options. But I'm happy to give default here..also This has more points on why providing defaults can be bad

[rtitle]

Yeah I don't love defaults either, was just curious. I don't really have a strong preference, will thumb as is :)

[rtitle]

Is this context not useful to log, or is it logged in another way?

[Qi77Qi]

when we log the PubsubMessage in the message, these context are already being logged. So it's a bit duplicate