Add ABIDES arena

[Muhtasham]

Summary

• add an ABIDES financial-market simulation arena backed by abides-sim/abides
• pin upstream ABIDES to c4bf157678928934417aba6073eb0651aeaf6d15, constrain Python dependencies, and pin pip in the arena image
• add the runtime adapter, starter abides_agent.py, example config, arena docs, and unit coverage

Runtime behavior

• evaluates each CodeClash player in its own identical seeded ABIDES market world per sim, with an exchange, market maker, and background zero-intelligence traders
• compares players by average mark-to-market profit across the configured market seeds
• computes scores from the player object holdings and final exchange price rather than mutable ABIDES agent.type / kernel.meanResultByAgentType accounting
• guards player lifecycle callbacks so one crashing submission receives CRASH_SCORE without crashing other players
• reloads player modules per sim, discards simulator stdout to /dev/null, and isolates ABIDES summary logs in temporary directories so runtime artifacts do not leak into the repo workspace

Hardening

• fixed validation-order bypass by importing the real ABIDES TradingAgent before executing the submitted module
• penalized missing player scores with CRASH_SCORE instead of treating absent ABIDES results as neutral 0.0
• changed from shared-kernel submitted-agent competition to isolated per-player worlds to avoid direct opponent-object access through self.kernel.agents
• added regression coverage for validation heredoc ordering and missing score handling

Verification

• uv run pytest tests/arenas/test_abides.py -> 7 passed
• uv run pytest tests/arenas -> 190 passed
• uv run pytest -> 192 passed
• uv run mkdocs build
• uv run pre-commit run --all-files
• uv run python main.py configs/examples/ABIDES__dummy__r1__s2.yaml -o /tmp/codeclash-abides-smoke-review-1777981176
  • rebuilt codeclash/abides from scratch
  • both players passed validation
  • produced round scores, e.g. -1736.0 vs -2297.0 depending on player ordering
  • produced metadata.json, game.log, tournament.log, and rounds/round_0.tar.gz / rounds/round_1.tar.gz
• manual runtime crash-isolation check with one agent raising in wakeup gave the bad agent -1000000.0 while the good agent still scored normally
• test ! -e log after local runtime/smoke checks confirmed no ABIDES log artifact leaked into the repo root