v1.0.7 - Dynamic models, reasoning history, and stream timeout fixes

Added

• Dynamic OpenCode Go model catalog from https://opencode.ai/zen/v1/models, cached for one hour in ~/.opencode-go-cli/opencode-models.json.
• --refresh-models to force a catalog refresh, standalone or with --list.

Changed

• opencode-go --list now shows the live OpenCode Go catalog while OpenAI, Qwen, and Z.ai keep their static curated lists.
• Expanded the built-in OpenCode Go fallback model snapshot from 5 to 13 entries.

Fixed

• Preserves Anthropic thinking and redacted_thinking history as OpenAI-compatible reasoning_content for Chat Completions providers.
• Adds a non-empty reasoning placeholder on assistant tool-call messages when upstream reasoning backends require it.
• Raises Bun proxy idleTimeout to 255 seconds so long upstream streams are not cut off by Bun's 10-second default.

Package

• Package version is now 1.0.7.
• CHANGELOG.md is included in the npm package files.
• Scoped package publish config is set to public access.

Verification

• bun run typecheck - clean
• bun test - 101 pass, 0 fail
• bun run build - clean
• bun pm pack --dry-run - package contains package.json, CHANGELOG.md, README.md, and dist/index.js

Full changelog: https://github.com/Lordymine/opencode-go-cli/blob/v1.0.7/CHANGELOG.md