feat(airplay): send the configured display name to the receiver

[nikhilshastry2003]

display_name in config.toml (and --name) only ever reached the sender's own window. Every AirPlay request — GET /info, POST /stream, in both the plain and the authenticated flow — sent the constant "OpenPlay" as X-Apple-Device-Name, which is the name a receiver shows for the sender. So a user who named their machine got "OpenPlay" on the TV regardless.

What changes

• http_session.rs — the OPENPLAY_DEVICE_NAME constant becomes pub const DEFAULT_DEVICE_NAME, and negotiate, get_info, post_stream, post_stream_on, get_info_raw and build_stream_request take a device_name: &str. Each runs it through a new header_safe_device_name() before putting it in the header: control characters (line breaks included) are dropped, whitespace trimmed, and a name with nothing left falls back to the default. A header ends at the first CR/LF, so without that a name from a hand-edited config could cut the request short and have its tail read as further headers. Non-ASCII is kept — Apple's own senders send it.
• session.rs — AirPlaySession::start gains device_name: &str. Internally the format, session id and name travel in a private SessionParams struct, which keeps run_session and negotiate_with_auth under clippy's seven-argument limit instead of adding an #[allow].
• openplay-sender — start_airplay_cast takes a CastSettings { bitrate_kbps, framerate, force_sw_encode, display_name } built from AppConfig in app.rs, for the same argument-count reason. run_airplay_pipeline stops taking width/height separately — they were already in the CaptureConfig it receives. Miracast's entry points are left as they are; they can move to the same struct when they need a fifth value.
• Examples — pair_probe and control_probe pass DEFAULT_DEVICE_NAME explicitly.
• Docs / CLI help — docs/configuration.md, the README usage block and the --name help text now say the name is sent to AirPlay receivers, not only shown in the window.

Tests

Four new unit tests in http_session.rs:

• the configured name appears as X-Apple-Device-Name in the built POST /stream
• a name containing \r\n cannot introduce a second header (the negative assertion fails if the sanitiser is removed)
• blank / whitespace-only / control-only names fall back to DEFAULT_DEVICE_NAME
• surrounding whitespace is trimmed and non-ASCII kept

Verified

• cargo fmt --all -- --check clean
• cargo clippy -p openplay-airplay --all-targets --all-features -- -D warnings clean
• cargo test -p openplay-airplay — 46 + 27 passed (was 42 + 27)

Not verified locally: openplay-sender. The machine this was written on has no GStreamer, so the three sender files (app.rs, casting.rs, main.rs) are compiled by CI only. The diff there is small and mechanical, but I want to state the gap rather than imply it was built.

Relation to #44

This touches the same negotiate / negotiate_with_auth region of session.rs that #44 rewrites (the retry-on-status block), so whichever lands second will need a small rebase. Happy to do that here once #44 merges — the overlap is the argument list on two calls, nothing in the logic.