Add MiniMax-M3 and MiniMax-M2.7 model configuration

[octo-patch]

Reason: The SWE-bench XML benchmark config and the model quickstart docs only referenced the older minimax/minimax-m2, with no configuration for the newer MiniMax-M3 or MiniMax-M2.7 models.

Changes

• src/minisweagent/config/benchmarks/swebench_xml.yaml: update the OpenRouter model used by the executable benchmark configuration from minimax/minimax-m2 to the current flagship minimax/minimax-m3.
• docs/models/quickstart.md: replace the single MiniMax example with two OpenRouter examples so both current models are documented — minimax/minimax-m3 and minimax/minimax-m2.7.

Both continue to use the existing openrouter model class and the same temperature: 0.0 model kwargs already used for MiniMax, so no code paths change.

Checks

• python3 -c "import yaml; yaml.safe_load(open('src/minisweagent/config/benchmarks/swebench_xml.yaml'))" — the benchmark config still parses and resolves model_name/model_class.

[klieret]

hi, thx, I just want to avoid too much clutter, I'm ok with adding one of them