Opinionated pre-selections for claude code subagents models

Latest version of Claude-Code included a support for model setting for each agent:
"Agents: Added model customization support - you can now specify which model an agent should use"
[Changelog](https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md)

using a right model for different agents not only can help with **costs/ limits** but also make development cycles more **faster** aka. not trying to use best and slowest model for everything.

aka orchestrator, planner needs reasoning (opus), coder or test runner can use (sonnet). Maybe documentation could use even haiku.
So add opinionated pre-selection of model for each role/agent