feat: add embedding APIs, batching, and docs

[leehack]

Summary

• add first-class embedding support in LlamaEngine with optional backend capability interfaces and native backend implementations (embed and embedBatch)
• introduce native worker/service embedding paths with multi-sequence batching controls (ModelParams.maxParallelSequences) plus benchmark/sweep tooling for throughput analysis
• add embedding example CLI, refresh README/example/website docs, and update template parity mapping for new vendored llama.cpp template fixture

Validation

• dart analyze
• dart test -p vm -j 1 --exclude-tags local-only
• dart test (in example/basic_app)
• npm run build (in website)

[codecov-commenter]

Codecov Report

❌ Patch coverage is 29.79592% with 172 lines in your changes missing coverage. Please review.
✅ Project coverage is 76.81%. Comparing base (d0734a0) to head (52d156c).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
lib/src/backends/llama_cpp/llama_cpp_service.dart	7.18%	168 Missing ⚠️
lib/src/backends/llama_cpp/llama_cpp_backend.dart	92.59%	2 Missing ⚠️
lib/src/backends/llama_cpp/worker.dart	83.33%	2 Missing ⚠️

❌ Your patch status has failed because the patch coverage (29.79%) is below the target coverage (70.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #75      +/-   ##
==========================================
+ Coverage   76.48%   76.81%   +0.33%     
==========================================
  Files          66       66              
  Lines        8338     8579     +241     
==========================================
+ Hits         6377     6590     +213     
- Misses       1961     1989      +28     

Flag	Coverage Δ
unittests	76.81% <29.79%> (+0.33%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[leehack]

Coverage update after backend/worker/model-params test expansion:

• Fixed a platform-sensitive assertion in llama_cpp_service_test (backend info now asserts CPU presence instead of exact list).
• Added and expanded unit coverage in backend routing, worker routing, and model params paths.
• Validation: dart analyze, targeted tests, and full VM suite (dart test -p vm -j 1 --exclude-tags local-only) all pass.
• Refreshed VM coverage (coverage/lcov.info). Current file coverage:
  • lib/src/backends/llama_cpp/llama_cpp_service.dart: 42.55% (680/1598)
  • lib/src/backends/llama_cpp/llama_cpp_backend.dart: 94.47% (239/253)
  • lib/src/backends/llama_cpp/worker.dart: 90.98% (111/122)
  • lib/src/core/models/inference/model_params.dart: 92.31% (12/13)
• Overall lib coverage now 76.27% (6544/8580).

[leehack]

Web bridge update completed:

• Published new bridge asset release: leehack/llama-web-bridge-assets@v0.1.6 (llama.cpp b8157).
• Updated llamadart web defaults/pins/docs to v0.1.6:
  • scripts/fetch_webgpu_bridge_assets.sh
  • example/chat_app/web/index.html
  • doc/webgpu_bridge.md
  • website/docs/platforms/webgpu-bridge.md
  • website/versioned_docs/version-0.6.4/platforms/webgpu-bridge.md
  • CHANGELOG.md (Unreleased note)
• Validation: dart analyze passed in llamadart.

Reference release: https://github.com/leehack/llama-web-bridge-assets/releases/tag/v0.1.6

[leehack]

Upstream tracking for the web bridge bump:

• Bridge source PR: build: bump web bridge llama.cpp pin to b8157 llama-web-bridge#1
• Published assets tag: https://github.com/leehack/llama-web-bridge-assets/releases/tag/v0.1.6

[leehack]

Web embeddings are now wired and validated.

What changed in this branch:

• Added web backend embedding support (LlamaEngine.embed / embedBatch) via WebGPU bridge APIs.
• Updated web backend wrappers/capabilities so embeddings resolve correctly on web.
• Added browser unit coverage for web embeddings and legacy bridge fallback behavior.
• Bumped default bridge assets to leehack/llama-web-bridge-assets@v0.1.7.

Validation run:

• dart analyze
• dart test -p chrome test/unit/backends/webgpu/webgpu_backend_test.dart test/unit/backends/web/web_backend_test.dart
• WEBGPU_BRIDGE_ASSETS_TAG=v0.1.7 ./scripts/fetch_webgpu_bridge_assets.sh (checksum verification passed)

Upstream references:

• Bridge PR: build: bump web bridge llama.cpp pin to b8157 llama-web-bridge#1
• Assets release: https://github.com/leehack/llama-web-bridge-assets/releases/tag/v0.1.7

[leehack]

Additional browser integration coverage added:

• test/integration/backends/webgpu/webgpu_engine_multimodal_browser_integration_test.dart
  • new assertion path for LlamaEngine.embed(...) and embedBatch(...) over the mock WebGPU bridge.

Re-validated:

• dart analyze
• dart test -p chrome test/unit/backends/webgpu/webgpu_backend_test.dart test/unit/backends/web/web_backend_test.dart test/integration/backends/webgpu/webgpu_engine_multimodal_browser_integration_test.dart