Trim repository Copilot instructions to reduce cloud agent session bloat

[Copilot]

The failing copilot Actions job was not failing on repo build or test logic; it was dying in a resumed Copilot cloud-agent session after accumulating a large instruction/prompt context. This change reduces repository instruction size while preserving the Fluent UI rules and validated commands the agent actually needs.

• Condense repository-wide Copilot guidance

  • Replaced a long, duplicated instruction file with a compact version focused on:
    • Nx/Yarn setup and command patterns
    • Fluent UI v9 component rules
    • expected file layout
    • targeted validation guidance
    • a few repo-specific guardrails

• Keep only high-signal repo context

  • Preserved guidance that affects agent correctness in this repo:
    • use Nx for tasks
    • prefer v9 patterns
    • avoid v8 copy/paste
    • use theme tokens
    • avoid direct DOM globals
    • avoid cross-package component dependencies

• Clarify instruction scope

  • Tightened wording around when to run install steps
  • Narrowed the dependency/security-instructions cross-reference to dependency/automation changes only

Example of the new instruction style:

## Validated setup and commands

- Run `yarn install --frozen-lockfile` during initial setup and after dependency changes.
- Always run project tasks through Nx:
  - `yarn nx run <project>:build`
  - `yarn nx run <project>:test`
  - `yarn nx run <project>:type-check`