perf(startup): reduce desktop first-screen work

[limityan]

Summary

Tracking issue: #949

• Show the desktop main window from the native startup path, while keeping the frontend show path as a fallback and visibility checkpoint.
• Move non-critical startup work behind the first visible shell / idle boundary, including IDE, MCP, ACP startup probes and MiniApp catalog refresh.
• Page session metadata loading so startup fetches only initially visible rows; explicit expand / show-more actions load the next page with loading feedback.
• Keep workspace and session sections visually expanded by default, so this does not hide existing navigation content to obtain the startup win.
• Remove the pre-React static logo layer and preserve the Tauri-injected theme background to avoid the black-logo-to-white-logo startup flash.
• Keep editor/LSP startup lazy and show a first-open loading prompt for the longer editor initialization path.
• Keep paged session reads on the maintained session index fast path; the page API validates stale entries only for the returned page, while full-list/rebuild paths keep the broader index integrity check.
• Maintain the session metadata index count incrementally on writes/deletes so the startup fast path does not reintroduce an O(number of sessions) directory scan during normal metadata saves.

Performance Data

Benchmark method: Windows desktop release-fast cold launches, 10 runs per build. The comparison uses the same packaged WebView path (tauri.localhost) for baseline and stage 1. Baseline snapshot: main@9fa991f6; PR head: bf1163d0. The branch is rebased onto gcwing/main@8a93ed70; the newer base commits are i18n contract guardrails and remote-connect owner migration changes, not frontend startup changes.

P50 metric	Baseline	Stage 1	Delta
Native window show since process start	n/a	888.5 ms	n/a
main_window_shown	1504.7 ms	1101.8 ms	-402.9 ms (-26.8%)
start_application_end	1455.7 ms	1035.6 ms	-420.1 ms (-28.9%)
startApplicationDurationMs	614.7 ms	27.7 ms	-586.9 ms (-95.5%)
afterRenderDurationMs	584.1 ms	461.4 ms	-122.7 ms (-21.0%)
Primary startup API count	15.5	7.0	-8.5 (-54.8%)
Primary startup response bytes	12,891	7,699	-5,192 (-40.3%)
Session metadata loaded	187	27	-160 (-85.6%)
Max session metadata duration	175.1 ms	53.3 ms	-121.8 ms (-69.6%)
interactive_shell_ready	1538.9 ms	1628.4 ms	+89.4 ms (+5.8%)

Additional notes:

• mainWindowShownReason was startup-native in all 10 stage-1 runs, with zero frontend fallback shows.
• The Windows maximize wait is preserved from the existing startup path. A previous no-wait run was excluded from the table because it was not an apples-to-apples packaged WebView run.
• A focused local metadata benchmark after the final audit fix reports page5_avg_ms=11.550 versus full_list_avg_ms=88.142 for 1000 sessions, or about 7.6x faster for the paged backend path.

Risks and Tradeoffs

• First visible window time improves by about 403 ms p50, but interactive_shell_ready p50 is still about 89 ms slower. Follow-up work should target React mount/effect cost and shell readiness directly.
• Deferred ACP / IDE / MCP startup means the first on-demand operation may pay the deferred fetch or startup cost. ACP menu client loading now has an explicit loading state.
• Session sections remain expanded by default. Only explicit user collapse suppresses metadata fetching, and clicking show more loads the next visible page.
• Paged session metadata now trusts the maintained index for startup performance and validates stale entries on the returned page. Out-of-band metadata writes outside the current page may require the full-list/rebuild path before they appear.
• Native startup now reveals a theme-colored shell before React renders. The extra pre-React logo was removed because it caused a double-logo flash.

Verification

• pnpm run fmt:rs
• pnpm run check:repo-hygiene
• pnpm run check:github-config
• pnpm run i18n:generate
• pnpm run i18n:audit
• pnpm run i18n:contract:test
• pnpm run lint:web
• pnpm run type-check:web
• pnpm --dir src/web-ui run test:run
• cargo test -p bitfun-core list_session_metadata_page -- --nocapture
• cargo test -p bitfun-core bench_session_metadata_page_vs_full_list -- --ignored --nocapture
• cargo test -p bitfun-core
• cargo check --workspace
• git diff --check
• git diff --cached --check