add Claude-specific effort level support

[eggfriedrice24]

Summary

• Add _get_effort_levels() to BaseProvider (nil) and ClaudeCodeProvider (low, medium, high, max)
• Add set_effort/get_effort public API with runtime validation
• Add select_effort() pickers for both telescope and fzf-lua
• Add tests for effort in _build_command and _get_effort_levels

Test plan

• make pr_ready passes (lint, tests, format)
• _99.set_effort("high") then trigger search — verify --effort high in command
• _99.set_effort("foo") — verify error notification
• _99.set_effort(nil) — verify effort cleared
• Switch to non-Claude provider, call set_effort — verify warning

---

Note

Medium Risk
Medium risk: changes provider request command construction and adds new public API state (set_effort/get_effort) that can affect execution for Claude runs; other providers are guarded to warn/ignore.

Overview
Adds Claude-specific “effort level” support end-to-end: a new provider hook BaseProvider._get_effort_levels() (default nil) and ClaudeCodeProvider._get_effort_levels() (low|medium|high|max), plus _99.set_effort()/_99.get_effort() with runtime validation and provider capability warnings.

Updates ClaudeCodeProvider._build_command() to optionally append --effort <level> when set, and introduces select_effort() pickers for both telescope.nvim and fzf-lua to let users choose effort interactively. Tests are expanded to cover effort flag inclusion/exclusion and the default nil behavior for non-Claude providers.

^{Written by Cursor Bugbot for commit c8168ec. This will update automatically on new commits. Configure here.}