Alt-backend hook so another server can answer Client calls

[tiper]

Before: plx_plex::plex::Client can only talk to a Plex server, so an app that wants to reuse PlxNative's screens, player and stores against another kind of media server has to fork the data layer.

After: behind a new alt-backend cargo feature (off in every PlxNative build, so nothing here is compiled into PlxNative), an application can install a plex::backend::Backend at boot, and the data, playback and live-update methods of Client hand their call to it and return its answer in the same model types. Without the feature, the alt! hook at the top of each method expands to nothing.

How: Backend is a trait whose every method defaults to "nothing here" (None, false, an empty path), so a Plex-only feature such as Discover or agent subtitle search simply disappears for a server without one. While a backend is installed:

• plex.tv and the Discover provider are never contacted (calls go to a closed local port, the existing offline path), and the account's audio and subtitle preferences are read from and saved to the backend.
• grant::PlaintextAsk::settle_vouched lets the backend's own sign-in mint a plaintext grant, under the same origin rule as mint (a numeric private-network literal, current generations, no refusal since). servers queues an endpoint re-proof (grant::strand) for a plaintext origin registered without a grant, since there is no plex.tv discovery to do it.
• the backend can name the app (identity::app_name, used by whatever shows a name to the viewer; PRODUCT stays what a Plex server is sent) and list its own credential spellings, so http::carries_credential treats them like X-Plex-Token.
• has_plex_home, rows_resolve_by_key, fixed_hls_available and versions give the app side the few answers that differ from Plex's.

Nothing changes for PlxNative builds.