fix(navigate): short-circuit FF152 events wait; honest elapsed_ms + committed_url

[ractive]

Summary

• On Firefox 152 the dom-complete document-event never fires for some static pages and SPAs, so a default navigate burned the full ~7s events budget before the readystate fallback rescued it (example.com ~7.26s vs --no-wait 0.06s). Theme A: the default both strategy now interleaves a lightweight, freshness-guarded document.readyState probe (ReadyStateProbe + probe_readystate_complete) into the wait_for_doc_complete drain loop and returns as soon as the page reports complete. Pages that do fire dom-complete promptly keep the richer event path (300ms head start, 250ms probe cadence).
• Theme B (timing): the single navigate-start Instant (nav_start) is threaded through both the events and readystate phases so elapsed_ms reflects true wall-clock instead of only the ~1ms readystate-poll duration.
• Theme B (URL): when a committing document-event carries no url (SPA case), committed_url is resolved via window.location.href (eval_location_href) rather than surfaced as about:blank.
• Docs: --wait-strategy help text, WaitStrategy::Both / WaitAfterNav doc-comments, and a new kb/rdp/actors/watcher.md note documenting the FF152 quirk and mitigation.

Test plan

• unit_navigate_readystate_probe_short_circuits — probe returns a complete CommitInfo with committed_url from location.href when no dom-complete ever arrives, well inside the events budget (mock server).
• unit_navigate_dom_complete_empty_url_falls_back_to_href — empty-URL dom-complete back-fills committed_url from location.href, never empty/about:blank.
• live_navigate_default_fast — default navigate returns in wall-clock < timeout/2.
• live_navigate_elapsed_matches_wall — results.elapsed_ms within ±750ms of measured wall-clock.
• live_navigate_spa_committed_url — committed_url is the real landed URL, never about:blank/empty.
• cargo fmt / cargo clippy --workspace --all-targets -- -D warnings / cargo test --workspace -q all clean.
• cargo xtask check-iteration-ready — 10/10 discipline gates pass (with FF_RDP_LIVE_TESTS=1).

🤖 Generated with Claude Code## Claims vs code
<generated 2026-07-19T11:36:11Z by ralph-loop>

• dom-complete → ✅ matched in diff
• dom-loading → ✅ matched in diff

[ractive]

Reviewer request (orchestrator): the three AC checkboxes name live tests but the PR shows no live-run output. Before merging, please run the navigate live suite against real FF152 and paste the result here:

FF_RDP_LIVE_TESTS=1 cargo test -p ff-rdp-cli --test live_navigate_default_fast -- --include-ignored

Expected: live_navigate_default_fast, live_navigate_elapsed_matches_wall, live_navigate_spa_committed_url all PASS. If any fail or Firefox can't be launched, do not merge — report instead.

[ractive]

Local code review (medium effort, local-only mode)

Ran a local review over git diff main...HEAD. Copilot review was skipped per invocation mode.

#	File:Line	Issue	Source
1	crates/ff-rdp-cli/src/commands/navigate.rs:369 (pre-fix)	Interleaved readystate probe ignored wait_level, so --wait loading/--wait interactive + --wait-strategy both could short-circuit with ready_state: "complete" instead of resolving on the requested dom event	Local
2	crates/ff-rdp-cli/src/commands/navigate.rs:279-287 (pre-fix)	WaitLevel::Loading early-return skipped the Theme B eval_location_href fallback applied to Interactive/Complete, so --wait loading could still emit an empty/about:blank committed_url for SPAs — the exact bug Theme B claims to fix	Local

Both fixed in 428f888: probe now gated to wait_level == WaitLevel::Complete, and the Loading path applies the same URL fallback as Interactive/Complete. Added unit_navigate_probe_ignored_for_non_complete_wait_level regression test. cargo fmt, cargo clippy --workspace --all-targets -- -D warnings, and cargo test --workspace -q all pass after the fix.