feat(opencode): route Responses models natively

[dgokeeffe]

Summary

• add OpenCode's native Databricks OpenAI/Responses provider
• bucket managed GPT/Responses models into that provider instead of silently discarding them
• select provider-qualified native model IDs in generated OpenCode configuration
• cover provider-specific routing and managed-model behaviour

Why

OpenCode was treating native GPT models as unsupported or routing them through a less appropriate compatibility path. That could hide models selected by workspace policy or produce gateway-incompatible requests. This gives native Responses models an explicit, tested provider boundary.

Stack

PR E of five. Depends on PR C (pr/capability-discovery). It is opened as a draft until A, B and C land; the isolated review delta is 763fa3e..210312e.

E and D are siblings and can be reviewed or landed independently once C is available.

• A: #446 — Pi routing foundation
• B: #447 — Claude discovery hardening
• C: #448 — capability-driven discovery
• D: #449 — sibling Pi OSS PR
• E: #450 — this PR

Validation

• 731 focused managed-resolution/OpenCode/CLI/discovery tests passed
• Ruff lint and formatting passed
• ty check src and git diff --check passed
• fresh installed-build OpenCode inference passed on system.ai.gpt-5-6-luna
• generated configuration used the native OpenAI provider and retained OSS setCacheKey: false
• the complete integration stack passed 2,358 tests with 37 skips; its three failures reproduce unchanged on main

Non-goals

This PR does not change shared model classification or Pi's streaming proxy.