Implement complete Rust scaffolding for LSP-powered context assembler

[Copilot]

Implements the full project scaffolding for skry as specified in README.md - a deterministic LSP client that builds LLM context by traversing dependency graphs.

Structure

• config - TOML parsing for token budgets, pruning strategies, LSP server mappings
• lsp - LSP client with subprocess lifecycle management
• pipeline - Intent parsing (symbol/search), entry point resolution, graph traversal
• assembly - Context stitching with token budgeting and configurable pruning
• standby - Daemon mode infrastructure for persistent LSP servers
• main - CLI with clap supporting generate, standby, query commands

Implementation Notes

• All LSP communication, IPC, and traversal logic are placeholder implementations marked with TODOs
• All pruning strategies (aggressive, bodies, and full) are placeholders with TODOs for language-specific AST parsing and logic detection
• Daemon query creates new instance rather than connecting via IPC (needs Unix socket/named pipe)
• 13 unit tests cover module interfaces and configuration parsing

Example Usage

# One-shot symbol resolution
skry generate --file src/main.rs --symbol ServerConfig --intent "Refactor to singleton"

# Search with depth limit
skry generate --search "auth_middleware" --depth 2

# Daemon mode
skry standby &
skry query --symbol "User::login"

Configuration via .skry.toml:

[general]
token_budget = 4096
pruning_strategy = "aggressive"

[lsp]
rs = "rust-analyzer"
cpp = "clangd"
ts = "typescript-language-server --stdio"

Original prompt

Implement the full, complete scaffolding for the project based on design described in README.md

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.