[Feature]: Data-driven Beats output for additional output types (Kafka/Redis/File/Console)

The Beats role only supports logstash and elasticsearch outputs (enforced via the beats_<beat>_output choices). Beats also supports Kafka, Redis, File and Console, which the role cannot render.

Make additional output types data-driven: a beats_<beat>_output dict (type + type-specific settings) rendered via a shared "dict → YAML" helper (same helper as the #393 inputs design). Elasticsearch and Logstash stay managed as today — the role keeps auto-wiring their hosts, credentials, TLS and setup, which a raw passthrough would lose.

Output-type conditions (e.g. setup gated on == "elasticsearch") then key off beats_<beat>_output.type; loadbalance folds into the output dict.

Note: Beats allows only one output, so "more outputs" means more output types, not several at once.

Breaking (changes the beats_<beat>_output variable shape). Related: #393, #512, #519.

Depends on: #393 (shared plugin-as-data macro).

Start with the beats__output choices and the shared plugin-as-data macro from #393. Trace the existing managed Elasticsearch and Logstash wiring, then verify that additional output types render through the shared dict-to-YAML helper, setup conditions use output.type, and loadbalance is included in the output dict.