feat(server): forward-port publish_raw_event_to + subscriber_addresses onto 0.8.0

[JustinKovacich]

Forward-ports the two targeted-delivery EventPublisher methods released in v0.7.2 (8118fde) onto the current 0.8.0 spine tip (feat/polled-port-onto-125), so the 0.8.0 release doesn't regress them.

Base corrected: originally opened against feature/phase18f_docs_and_0_8_0_finalize; retargeted to the real spine tip feat/polled-port-onto-125 and re-authored against the post-#125 caller-scratch publish API.

What

• publish_raw_event_to_with_buffers(target, …, buf) — sends a raw (already-serialized, already-E2E-protected) event to one subscriber by endpoint, taking caller-provided scratch. Same shape as the post-Reduce async state and static pool memory in simple_someip to prevent Embassy arena exhaustion #125 publish_raw_event_with_buffers fan-out (guards keyed off buf.len(), the buf.len() < 16 floor, socket via SharedHandle::get()).
• publish_raw_event_to(target, …) — #[cfg(feature = "_alloc")] convenience wrapper that allocates UDP_BUFFER_SIZE scratch and delegates, mirroring publish_raw_event.
• subscriber_addresses(…) — lists an event group's subscribed endpoints into a heapless::Vec<_, SUBSCRIBERS_PER_GROUP> (no_std-clean).

Enables per-recipient delivery when subscribers share a (service, instance, event_group) key but bind distinct endpoints (multiple sensors on one subnet sharing a fixed instance id). Additive; existing fan-out unchanged.

One deliberate semantic change vs v0.7.2

A send failure to a subscribed target returns Err(Error::Transport(_)) rather than the v0.7.2 Ok(0) — matching the post-#125 fan-out path, so callers can distinguish "not subscribed" (Ok(0)) from "subscribed but send failed" (Err).

Verification

• cargo fmt -- --check — changed file is clean ✓
• cargo clippy --features server-tokio --tests — changed file is warning-clean ✓
• cargo nextest run --features server-tokio event_publisher — 23/23 pass (incl. 3 new) ✓
• no_std: cargo build --target thumbv7em-none-eabihf --no-default-features --features server,bare_metal and client,server,bare_metal ✓

Heads-up (pre-existing, not from this PR): the base tip already fails cargo fmt --all --check under rustfmt 1.9.0 (~10 drifted files: runtime.rs, sd_state.rs, client/*, etc.) and cargo clippy -Dwarnings (nits in registry.rs, lib.rs, buffer_pool.rs, …). Those are untouched by this change; this PR's fmt/clippy CI may go red on them until the base is cleaned.

🤖 Generated with Claude Code

[JustinKovacich]

Folded into #138. The 0.8.0 forward-port stack was consolidated via a linear rebase of the whole stack onto main (which now includes #130). #138 is rebased onto main and carries this branch's commits intact — closing here to collapse the stack into a single PR. Verified: 556 all-features lib tests, client_server 11/11, all 5 CI clippy lanes, no_alloc_witness, bare_metal_e2e 9/9.