🚀 Release v2.5.1 — configure inline completion endpoints (#132)

## Change Log / Overview

### 🚀 Features
- Add optional provider-group `completionsUrl` configuration for
OpenAI-compatible FIM endpoints.
- Preserve configured LiteLLM paths such as `/v1` when deriving
`/completions`.
- Support model-level `completionsUrl` overrides with model → group →
derived endpoint precedence.
- Publish resolved endpoint metadata for downstream VS Code
inline-completion support.

### 🧪 Tests
- Add endpoint derivation, path preservation, precedence, and invalid
URL fallback coverage.

### 📚 Documentation
- Update `README.md`, `README.marketplace.md`, and `CHANGELOG.md` for
v2.5.1.

### Related Issues
- #130

### Pull Request Pre-check
- [x] Linting Validation Passed (`npm run lint`; existing warnings only)
- [x] Formatting Validation Passed (`npm run format`)
- [x] Unit Testing Passes (`npm run test:coverage`)
- [x] Documentation Updated
- [x] Compile Validation Passed (`npm run compile`)

> Note: Inline completion behavior still depends on upstream VS
Code/Copilot support for contributed-provider `completionsUrl`
consumption.