fix(core): correct turn-latency metrics and idle-window bg status visibility

[bug-ops]

Summary

• MetricsBridge::WATCHED_SPANS watched the bare llm.chat span instead of llm.chat_with_tools, the span every tool-enabled turn (essentially all of them) actually dispatches through — the turn-latency panel showed llm:0ms even when the real LLM call took 25+ seconds. Auxiliary bare llm.chat calls in the same turn (self-check, compaction probe, etc.) were also silently overwriting the correct manually-tracked duration.
• Fixed by watching llm.chat_with_tools, accumulating across a turn's multi-round tool loop (instead of overwriting), and scoping accumulation to spans nested under the main turn's llm.turn_call span so concurrent in-process sub-agent and scheduler chat_with_tools calls can't inflate or corrupt the main turn's llm_chat_ms (found during review).
• The bg: N enrich, M telem TUI status segment only refreshed at the start of the next turn, so it stayed stale/invisible during the entire idle window after a response — exactly when background enrichment/telemetry extraction is actually running. Added a periodic BgMetricsTick to the existing Agent::next_event tokio::select! loop (no new tokio::spawn), using interval_at to defer the first tick and avoid racing the pre-existing channel-closed shutdown path (found during testing).

Closes #6275
Closes #6279

Test plan

• cargo +nightly fmt --check
• cargo clippy --profile ci --workspace --all-targets --features "desktop,ide,server,chat,pdf,scheduler,testing" -- -D warnings
• cargo clippy --profile ci -p zeph-core --all-targets --features "profiling,scheduler" -- -D warnings (metrics_bridge.rs is profiling-gated, not covered by the CI feature matrix)
• cargo nextest run --config-file .github/nextest.toml --workspace --features "desktop,ide,server,chat,pdf,scheduler" --lib --bins (13636 passed)
• RUSTFLAGS="-D warnings" RUSTDOCFLAGS="--deny rustdoc::broken_intra_doc_links" cargo doc --no-deps --workspace --features "desktop,ide,server,chat,pdf,scheduler"
• New/updated regression tests in metrics_bridge.rs, agent/utils.rs, agent/tests/bg_metrics_tick_race_tests.rs, agent/tests/agent_tests/lifecycle_tests.rs
• Live --tui session confirmation per updated playbooks (metrics-bridge-watched-spans.md Scenario 2/2b/2c, bg-supervisor-phase2.md S9) — deferred to next live-testing cycle, coverage-status.md rows left Partial