fix(openai): support gpt-5.6 reasoning_effort with function tools

[vishal27shetty]

Problem

gpt-5.6, gpt-5.6-luna, gpt-5.6-sol, and gpt-5.6-terra aren't in ChatModels or
_supports_reasoning_effort(), so the plugin doesn't recognize them as reasoning models at all.

This causes two independent 400s from OpenAI. First, an unspecified reasoning_effort falls
through to the "minimal" default, which this family rejects outright, tools or not:

Unsupported value: 'reasoning_effort' does not support 'minimal' with this model.
Supported values are: 'none', 'low', 'medium', 'high', 'xhigh'.

Second, once function tools are attached, gpt-5.6 requires reasoning_effort to be explicitly
"none" — omitting the parameter also 400s, unlike gpt-5.2/gpt-5.4, which just need it absent:

Function tools with reasoning_effort are not supported for gpt-5.6-luna in
/v1/chat/completions. To use function tools, use /v1/responses or set
reasoning_effort to 'none'.

Changes

• models.py: register the 4 gpt-5.6 variants in ChatModels and _supports_reasoning_effort;
  add _REASONING_EFFORT_NONE_MODELS, the set of models whose correct default is "none".
• llm.py: use _REASONING_EFFORT_NONE_MODELS for default selection (previously a hardcoded
  list missing gpt-5.6); add _REASONING_EFFORT_TOOL_REQUIRES_NONE_MODELS and force
  reasoning_effort="none" at request time when tools are attached, scoped to gpt-5.6 only.
• responses/llm.py: same default-selection fix — the Responses API already accepts tools with
  any reasoning_effort value, verified live, so no coercion is needed there.

Testing

Live against the OpenAI API across all 8 affected models (gpt-5.1, gpt-5.2, gpt-5.4,
gpt-5.4-mini, and all 4 gpt-5.6 variants), with function tools attached, both with no
reasoning_effort set and with an explicit override — 16/16 passing. Confirmed
_REASONING_EFFORT_TOOL_REQUIRES_NONE_MODELS contains only the 4 gpt-5.6 models by inspecting
it directly at runtime; gpt-5.1/5.2/5.4/5.4-mini are unaffected.

[CLAassistant]

All committers have signed the CLA.

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.