Add terminal command entrypoints for LSPRAG workflows

[Huang-Xuwei]

Summary

This PR adds terminal-accessible entrypoints for core LSPRAG workflows, so they no longer depend on opening the Command Palette in VS Code.

Specifically, this change introduces CLI support for:

• testing LLM connectivity
• generating unit tests from a file/function target
• printing the current LSPRAG configuration

These CLI commands are designed to reuse the existing extension behavior as much as possible, instead of creating a separate implementation path.

Motivation

LSPRAG previously exposed its main workflows only through VS Code commands such as:

• LSPRAG: Test LLM
• LSPRAG: Generate Unit Test
• LSPRAG: Show Current Settings

That made scripting, automation, remote usage, and terminal-first debugging harder than necessary. This PR adds a terminal path for those workflows so they can be invoked directly from the workspace root.

Changes

New CLI commands

Added terminal commands for:

• npm run test:llm
• npm run generate:cli -- --file <path> --function <name>
• npm run show:settings

Shared command logic

Refactored command implementations so the CLI path reuses existing extension logic where possible.

• generate:cli reuses the extension command path instead of introducing a separate generator implementation
• show:settings and the VS Code command now share the same settings summary formatter

CLI host and VS Code test integration

For generate:cli, the PR adds a VS Code test-host bridge so terminal execution can still use:

• document symbols
• semantic tokens
• language server features
• the existing generation pipeline

WSL / Remote robustness improvements

The terminal generation path now handles several environment-specific issues that prevented the VS Code test instance from loading language extensions correctly:

• isolates the spawned VS Code process from VSCODE_* remote CLI state
• syncs required language extensions into the test profile
• clears stale .obsolete extension removal markers
• allows configurable VS Code test versions, defaulting to stable

Better CLI validation and errors

Improved terminal UX by failing earlier and with clearer messages:

• validates LLM configuration before launching the test host
• reports missing provider-specific credentials directly in terminal
• when a target function name is not found, prints available callable names in the file
• corrected misleading CLI examples to use actual callable symbols from the fixture

Files added / updated

Main additions

• src/cli/testLLM.ts
• src/cli/generateUnitTest.ts
• src/cli/showSettings.ts
• src/llmHealthcheck.ts
• src/currentSettings.ts
• src/commands/generateUnitTestCommand.ts
• src/test/generateCliHost.ts

Main updates

• src/extension.ts
• src/invokeLLM.ts
• src/config.ts
• src/test/runTest.ts
• src/test/runExperiment.ts
• package.json
• README.md

Usage

Examples:

npm run test:llm
npm run show:settings
npm run generate:cli -- --file src/test/fixtures/python/calculator.py --function compute

Validation

Validated locally:

• TypeScript compilation passes
• terminal settings command works
• terminal generate flow now reaches the extension/LSP pipeline correctly
• missing LLM credentials are reported before generation starts

Notes

Full end-to-end generation still requires a valid configured LLM provider, for example:

• DEEPSEEK_API_KEY
• OPENAI_API_KEY
• or LOCAL_LLM_URL for local/Ollama usage