fix(mistral): add devstral models with correct 262K context length

[Sean-Kenneth-Doherty]

Summary

Fixes #6185

Problem

Devstral models (devstral-2512 and devstral-small-2505) were missing from the Mistral provider's declarative model list, causing Goose to fall back to the default 128K context limit instead of the actual 262144 tokens supported by these models.

Solution

Added both devstral models to mistral.json with their correct context_limit values (262144) as confirmed by the Mistral API:

curl -s https://api.mistral.ai/v1/models -H "Authorization: Bearer $MISTRAL_API_KEY" \
  | jq '.data[] | select(.id == "devstral-2512") | {id, max_context_length}'
# Returns: {"id": "devstral-2512", "max_context_length": 262144}

Changes

• Added devstral-2512 with context_limit: 262144
• Added devstral-small-2505 with context_limit: 262144

Testing

• Verified the context values match Mistral's official API documentation
• The canonical_models.json already had these models with correct context limits for OpenRouter

[michaelneale]

thanks @Sean-Kenneth-Doherty - can you sign the DCO so we can get this in?

[michaelneale]

can merge when DCO signed, thanks!