Improved Datadog integration

[markbastiaans]

This PR includes the following changes to the Datadog and Telegraf integrations:

• Use Telegraf agent for collecting PostgreSQL metrics and database diskstorage size metric. Telegraf will send these metrics directly to Datadog if Datadog is enabled.
• Use Datadog IoT and trace agents for collecting application metrics, JMX metrics, logs and traces
• Use official Datadog Cloud Foundry buildpack. This saves 400+ MB of application disk space and makes the integration more maintainable in the future.
• Added logs redaction - email addresses will be redacted by default. Can be switched off by setting the DATADOG_LOGS_REDACTION environment variable to false.
• Removed direct Databroker dependencies in Datadog module
• Refactoring and improvements: tags, configuration templates and more

Known Limitations

Telegraf does not support (Datadog) metric types correctly yet (e.g. rate, counter, gauge). This means that all database metrics are currently pushed to Datadog as a gauge.

The most important metrics (before_xid_wraparound, connections, database_size, db.count, locks, max_connections, percent_usage_connections, table.count, deadlocks) are gauges and are compatible with the Datadog PostgreSQL integration and associated dashboards.

If you do require the additional rate and counter metrics, there is a workaround available. First, set the DATADOG_DATABASE_RATE_COUNT_METRICS environment variable to true. After that variable is enabled, the rate and counter metrics are suffixed by either _count or _rate to prevent collisions with the official Datadog metrics. In the Datadog UI, the metric type and unit can be changed to reflect this. We also set a helpful interval tag (10s) which can be used here. Additionally, gauge metrics can be rolled up in Datadog dashboards. The correct type and unit for other submitted metrics can be found here.

[matyaskuti]

Looks okay in general, a few small comments and questions, nothing big related to the overall goal.