Produce failed messages to dead-letter-queue for swatch-azure-producer

swatch-azure-producer CONSUMES messages from billable_usage topic

• Processing FAILS due to azureUsageContext not available yet.  using 72 hours from snapshot_date to in order to determine hard failure vs. retryable failure
• Make sure 72 isn’t a magic number.  A constant with java comment is ok.  Putting it as an app properties value that COULD be overridden is ok.  Don’t expose via env var/clowdapp.  The number is unlikely to change unless Azure does things that make the missing subscription id causing this mess available earlier.

• swatch-azure-producer PRODUCES messages to billable-usage.dlt regardless of hard vs retryable failures
• Preserve the original message, so we can reuse this DLQ for any billable-usage failures in the future
• Include message headers
• retry_after = now() + 1hr

• On hard failures that we know (such as we are after the 72 hour window since snapshot_date for update from Azure), errors should be logged at Warning level, with a descriptive log message & structured log data to enable debugging/alerting. On unknown hard failures ERROR with stack trace

AC:

• When swatch-azure-producer receives a billableUsage and the subscription can not be determined (subscriptionId unavailable), a message is produced to billable-usage.dlt with retry_after header set to now() + 1hr
• When the subscriptionId is unavailable and the usage is past 72 hours (configurable) old, log a warning
• Log non retriable other errors at either warning or error