Support for Databricks Foundation Model APIs provider

[dgokeeffe]

Add Databricks as a new provider to support Foundation Model APIs.

Problem:
Databricks users cannot use opencode with their Databricks workspace's Foundation Model APIs (Claude, Llama, GPT, Gemini models) which are available as pay-per-token endpoints.

Proposed solution:

• Add Databricks provider with OpenAI-compatible endpoint support
• Support multiple auth methods: PAT tokens, OAuth M2M, Azure AD Service Principal
• Include default model definitions for common Foundation Model API endpoints
• Add auth command guidance for Databricks authentication setup

Verification:

• Unit tests for provider configuration parsing
• Tests for auth precedence and URL handling

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• "INVALID_PARAMETER_VALUE: Gemini models only support one system prompt" when prompting a Gemini model through Databricks (OpenAI API compatible provider) #4699: "INVALID_PARAMETER_VALUE: Gemini models only support one system prompt" when prompting a Gemini model through Databricks (OpenAI API compatible provider) - indicates Databricks is already partially supported
• Stream option "include_usage" not supported for AWS Bedrock models through Databricks (OpenAI API compatible provider) #4698: Stream option "include_usage" not supported for AWS Bedrock models through Databricks (OpenAI API compatible provider) - shows Databricks integration exists with some limitations
• Auto-discover models from OpenAI-compatible provider endpoints #6231: Auto-discover models from OpenAI-compatible provider endpoints - related feature for auto-discovering models from OpenAI-compatible endpoints
• Add Unbound as a provider #7594: Add Unbound as a provider - similar request for adding a new provider with OpenAI-compatible support
• [FEATURE]: multiple auth profiles per provider #5391: [FEATURE]: multiple auth profiles per provider - relevant for the "multiple auth methods" aspect mentioned in this issue

Feel free to ignore if none of these address your specific case.