perf(native): improve inference hot paths and add parity tooling

[leehack]

Summary

• Reduce native inference overhead in SDK hot paths by caching metadata, making prompt-token counting optional for create(...), batching worker stream chunks, and adding prompt-prefix reuse with deterministic full-replay fallback.
• Optimize ChatSession context trimming with bounded turn-offset search, add configurable stream batching/reuse knobs in GenerationParams, and extend unit coverage for the new behavior.
• Add native benchmark and prompt-reuse parity tools, wire CI parity checks in ci.yml, and prepare the 0.6.2 release notes/version/doc snippets.

Validation

• dart analyze
• dart test
• dart run tool/testing/native_prompt_reuse_parity.dart --model "example/basic_app/models/qwen2.5-0.5b-instruct-q4_k_m.gguf" --prompt-file "tool/testing/prompts/native_prompt_reuse_parity_prompts.txt" --runs 2 --max-tokens 128 --stream-batch-tokens 8 --stream-batch-bytes 512 --fail-on-mismatch

[codecov-commenter]

Codecov Report

❌ Patch coverage is 93.10345% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 77.26%. Comparing base (b252507) to head (9505d29).
⚠️ Report is 4 commits behind head on main.

Files with missing lines	Patch %	Lines
lib/src/backends/llama_cpp/llama_cpp_service.dart	87.27%	7 Missing ⚠️
lib/src/core/engine/engine.dart	85.71%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #61      +/-   ##
==========================================
+ Coverage   76.91%   77.26%   +0.35%     
==========================================
  Files          65       66       +1     
  Lines        7930     8046     +116     
==========================================
+ Hits         6099     6217     +118     
+ Misses       1831     1829       -2     

Flag	Coverage Δ
unittests	77.26% <93.10%> (+0.35%)	⬆️

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.