docs: RabbitMQ connection and queue arguments pages, application lifecycle, raw bytes content-type

[Lancetnik]

Summary

Closes the documentation gaps that Context7 users hit most often. Each item is a question we had no page for; the answers were only in source or tests.

• RabbitMQ Connection (rabbit/connection.md): the AMQP URL vs. host / port / virtualhost / ssl_options keyword arguments and their precedence, virtual host handling, one broker per virtual host, and the difference between connect() and start(). declare.md now points at that section because declare_queue / declare_exchange need an open connection.
• RabbitMQ Queue Arguments (rabbit/queue_arguments.md): how to pass x-* arguments through RabbitQueue(arguments=...), a table of the common ones per queue type, the PRECONDITION_FAILED trap on redeclaration, and a full dead letter queue section with a runnable example (RejectMessage → DLX → DLQ, x-death header).
• Application Lifecycle (getting-started/lifespan/lifecycle.md): the contracts of run(), start(), stop() and exit(), who owns the signal handlers, how to embed FastStream in your own event loop, why two run()s in one process don't work, and how ASGI and TestApp fit in. One snippet per broker.
• Raw bytes and content-type (getting-started/publishing/index.md): bytes are sent without a content-type header; how to set one through headers= or, for RabbitMQ, the content_type= argument; what the consumer sees.

Test plan

• tests/docs/rabbit/test_connection.py — in-memory, checks the built URLs and the two-vhost application
• tests/docs/rabbit/test_dead_letter.py — connected, run against a local rabbitmq:3.13-alpine: the rejected message reaches the DLQ handler
• tests/docs/getting_started/lifespan/test_manual_run.py — in-memory for all six brokers, drives serve() as a task and stops it through the event
• ruff check / ruff format on the new test files
• mkdocs build --strict for the new pages and links

🤖 Generated with Claude Code