Spatial API Unification

Currently, mesa-llm manually checks for both legacy `pos` (from `mesa.space`) and `cell.coordinate` (from `mesa.discrete_space`) in multiple places.
This change introduces a unified spatial API within `LLMAgent`.

### Proposed Changes

**[MODIFY] `llm_agent.py`**

- Add `@property pos` that returns the correct coordinate by checking both `self.cell` and base `Agent.pos`.
- Add `move_to(pos)` method:
- Uses `grid.move_agent()` for `mesa.space`
- Updates `self.cell` for `mesa.discrete_space`
- Update `_build_observation()` to use `self.pos`.

**[MODIFY] `inbuilt_tools.py`**

- Remove `_get_agent_position` helper.
- Update `@requires` decorators to use `agent.pos`.
- Simplify `move_one_step` and `teleport_to_location` to use `agent.pos` and `agent.move_to()`.

Start with llm_agent.py and inbuilt_tools.py, then read the existing mesa.space and mesa.discrete_space handling around LLMAgent, _build_observation(), and the movement tools. Done means both spatial APIs use LLMAgent.pos and move_to(), the helper is removed, and observation and movement behavior remain correct for both cases.

Written by the indexing model from the issue text.