fix(cli): flush tracing guards on all remaining application-level exit sites

[bug-ops]

Summary

Follow-up to #6698/#6708 — closes the last application-level std::process::exit sites that bypassed TracingGuards::drop's flush-on-drop logic. The issue named two sites (src/bootstrap/config.rs:27, seven in src/commands/bench.rs), but review turned up 16 more instances of the same pattern reachable from run() after init_tracing.

Rather than threading a bespoke exit-signal type through each site, this unifies on plain anyhow::Result + ? propagation everywhere: load_config_or_default now returns Result<Config, ConfigLoadError> instead of exiting internally, and all 18 callers (including bench.rs's 7 sites) use ?/anyhow::bail! so an ordinary Err return unwinds and flushes tracing_guards via normal drop. No new machinery, less code than the original approach.

Verified: zero application-level std::process::exit calls remain in src/ outside tracing_init.rs's own sanctioned flush primitives.

Closes #6709

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 --workspace --all-targets --features bench -- -D warnings (separate bench matrix entry)
• cargo nextest run --config-file .github/nextest.toml --workspace --features "desktop,ide,server,chat,pdf,scheduler" --lib --bins — 15312/15312 passed
• RUSTFLAGS="-D warnings" RUSTDOCFLAGS="--deny rustdoc::broken_intra_doc_links" cargo doc --no-deps --workspace --features "desktop,ide,server,chat,pdf,scheduler"
• cargo test --doc --workspace --features "desktop,ide,server,chat,pdf,scheduler"
• 8 new tests covering the previously-untestable error branches (config parse failure, handle_url_open error path, 6 in bench.rs)
• gitleaks protect --staged --no-banner --redact