feat(orchestration): make whole-plan verify timeout independently configurable

[bug-ops]

Summary

• verify_plan() (whole-plan grounding) shared orchestration.verifier_timeout_secs with per-task verify(). Live testing (ci-1402/1403/1404) observed it consistently exhausting the full 120s budget and fail-opening (complete=true, confidence=0.0, gaps=0) while per-task verify() on the same run/provider returned promptly.
• Adds whole_plan_verifier_timeout_secs (0 = fall back to verifier_timeout_secs, mirrors EnsembleConfig::member_timeout_secs), wired into a new PlanVerifier::whole_plan_timeout field used by verify_plan()'s timeout call and both its timeout-path log fields.
• Adds tracing::debug! prompt-size/tool-trace-entry diagnostics to both verify() and verify_plan() right before their LLM calls, so a future live session can tell a prompt-size effect apart from a genuine model-latency/hang issue — root-cause analysis found the whole-plan prompt is not obviously larger than a per-task prompt in the 2-task repro case, so a longer timeout alone is an unproven fix.
• Adds config migration Step 93 (advisory comment under [orchestration]).

Scope note: default behavior is unchanged (0 still resolves to the same 120s), so this does not resolve the live symptom out-of-the-box on its own — it closes the config/spec gap that prevented giving verify_plan() a different budget at all, and adds the instrumentation needed to determine and confirm the actual fix in a follow-up live-verification session.

Refs #6379

Test plan

• cargo +nightly fmt --check
• cargo clippy --profile ci --workspace --all-targets --features "desktop,ide,server,chat,pdf,scheduler,testing" -- -D warnings
• cargo nextest run --config-file .github/nextest.toml --workspace --features "desktop,ide,server,chat,pdf,scheduler" --lib --bins (14211 passed)
• RUSTFLAGS="-D warnings" RUSTDOCFLAGS="--deny rustdoc::broken_intra_doc_links" cargo doc --no-deps --workspace --features "desktop,ide,server,chat,pdf,scheduler"
• New unit tests: config default/round-trip, PlanVerifier fallback/independence (construction-level + behavioral via SlowMockProvider), 4 migration Step 93 tests
• CHANGELOG.md updated ([Unreleased] > Fixed)
• Live LLM round-trip verification (out of scope for this session; tracked via updated .local/testing/coverage-status.md "Whole-plan verifier grounding" row, left Partial)