feat: strategic placement engines for AI players

[njbrake]

Description

Enrich setup prompts so LLM players can reason strategically about settlement placement during the setup phase. Previously, AI players received bare vertex coordinates like (0, -2, N) with zero information about adjacent resources, probability, ports, or spatial relationships. Now each legal vertex shows:

• Adjacent resources with number tokens (e.g. Ore(10), Sheep(2), Wood(9))
• Total pip count for probability reasoning
• Expansion potential -- how many adjacent vertices satisfy the distance rule for future settlements
• Opponent proximity -- other players' buildings on shared production hexes
• Own buildings -- player's existing settlements (round 2 connectivity context)
• Port access info (e.g. 2:1 Wheat port)

Example annotation:

  3. (0,-1,S) | Wheat(12), Wheat(11) | pips=4 | expand=2
  7. (2,-2,N) | Wheat(9), Brick(6) | pips=9 | expand=1 | 2:1 Wheat port | near: P1
 12. (-1,0,N) | Wood(9), Ore(8) | pips=9 | expand=3 | your settlements: (0,-2,N)

Also adds per-personality setup strategies injected into the prompt, and fixes the hardcoded round=1 bug in LlamafilePlayer.

Changes:

• board.rs: Add get_hex(), port_at_vertex(), public pip_count() helpers
• board_view.rs: Deduplicate probability_dots -> use board::pip_count
• prompt.rs: Add annotate_vertex() with resource, spatial, and expansion context; enrich setup_settlement_prompt with legend
• personality.rs: Add optional setup_strategy field + setup_strategy_text() method + built-in strategies for all 5 personalities
• mod.rs: Add round: u8 to Player::choose_settlement trait
• orchestrator.rs: Pass actual round number to choose_settlement
• llamafile_player.rs: Use real round, inject strategy + player names into prompt
• All other player impls: Accept new _round: u8 parameter
• diplomat.toml / rusher.toml: Add setup_strategy

Fixes #35

PR Type

• New Feature
• Bug Fix
• Refactor
• Documentation
• Infrastructure / CI

Checklist

• I understand the code I am submitting
• New and existing tests pass
• Documentation was updated where necessary

AI Usage

• No AI was used
• AI was used for drafting/refactoring
• This is fully AI-generated

AI Model/Tool used: Claude Opus 4.6 via Claude Code

Any Additional AI Details you'd like to share: Implemented all sub-tasks from the issue plus spatial context enrichment based on research into LLM board game representations (Diplomacy-style per-decision relational context). Added 13 new tests covering all new functionality.

• I am an AI Agent filling out this form (check box if true)