Add manual benchmark report pipeline with GitHub Pages publishing

[Copilot]

This change adds a manually triggered benchmark pipeline for comparing bitnet-b1.58-sharp and traditional-local on the same built-in training corpus and query script. It produces a publishable report that combines integration-style efficacy/accuracy results with BenchmarkDotNet performance output and deploys that report through GitHub Pages.

• CLI/report generation

  • Adds a benchmark-report command to BitNetSharp.App
  • Runs the existing BenchmarkDotNet suites for the selected models
  • Evaluates both models against the shared default query set from BitNetTrainingCorpus.CreateDefaultExamples()
  • Emits a static report bundle:
    • index.html
    • comparison-report.md
    • comparison-report.json
    • raw BenchmarkDotNet HTML/CSV/Markdown exports

• Report contents

  • Captures:
    • efficacy: non-empty response rate across the shared query script
    • accuracy: exact-match rate and expected-token recall against corpus responses
    • performance: BenchmarkDotNet host/query/stream/train measurements
  • Preserves the current model behavior:
    • BitNet remains inference-only
    • training benchmarks stay scoped to trainable models

• GitHub Actions / Pages

  • Adds a dedicated manual workflow at .github/workflows/benchmark-report.yml
  • Workflow:
    • builds/tests the solution
    • runs benchmark-report
    • uploads the generated report as an artifact
    • deploys the static site to GitHub Pages

• Tests

  • Adds focused coverage for:
    • parsing BenchmarkDotNet markdown tables into comparison rows
    • writing the static HTML/Markdown/JSON report site

• Docs

  • Documents the new benchmark-report command
  • Documents the manual benchmark-report workflow and generated artifact layout

Example:

dotnet run --configuration Release \
  --project src/BitNetSharp.App/BitNetSharp.App.csproj -- \
  benchmark-report \
  --model=bitnet-b1.58-sharp \
  --compare-model=traditional-local \
  --output=/absolute/path/to/benchmark-report

This produces a Pages-ready report directory with a top-level index.html plus linked raw BenchmarkDotNet results.

• Example output shape

  benchmark-report/
  ├── index.html
  ├── comparison-report.md
  ├── comparison-report.json
  └── BenchmarkDotNet.Artifacts/
      └── results/
          ├── *-report.html
          ├── *-report.csv
          └── *-report-github.md
  

Generated benchmark report UI:
https://github.com/user-attachments/assets/3ae2bfa5-0508-45f4-afd4-524fd6993420

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[Copilot]

Pull request overview

Adds a new “benchmark-report” entry point to the BitNetSharp.App CLI and CI to generate a static benchmark comparison site (HTML + Markdown + JSON) by combining BenchmarkDotNet exports with a shared corpus/query-script evaluation.

Changes:

• Introduces HostedAgentBenchmarkReportRunner to run benchmarks, parse BenchmarkDotNet markdown tables, and write HTML/Markdown/JSON report artifacts.
• Adds benchmark-report command wiring in Program.cs plus a manual GitHub Actions workflow to publish the report via GitHub Pages.
• Adds tests for markdown table parsing and report artifact generation; updates docs with usage instructions.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
tests/BitNetSharp.Tests/HostedAgentBenchmarkReportRunnerTests.cs	Adds unit tests for performance-row parsing and report site artifact generation.
src/BitNetSharp.App/Program.cs	Adds benchmark-report command handling to generate the comparison report output directory.
src/BitNetSharp.App/HostedAgentBenchmarkReportRunner.cs	Implements report generation (benchmark execution, corpus/query evaluation, parsing BDN exports, writing HTML/MD/JSON).
docs/usage.md	Documents the new benchmark-report command usage.
docs/benchmarking.md	Expands benchmarking docs to describe the comparison report site and GitHub Pages workflow.
.github/workflows/benchmark-report.yml	Adds a manual workflow that builds/tests, generates the report, uploads artifact, and deploys to Pages.