feat(aihubmix): add gemini-3.5-flash-lite

[0genlab]

Adds the missing AIHubMix route gemini-3.5-flash-lite, inheriting lab metadata from google/gemini-3.5-flash-lite and keeping only host-specific pricing, controls, and actual overrides.

Sources:

• AIHubMix public catalog, retrieved 2026-09-20; prices are USD per million tokens.
• Model/control documentation.
• Existing baseline: providers/google/models/gemini-3.5-flash-lite.toml.
• AIHubMix request field documentation.

Reasoning controls: [{"type":"effort","values":["minimal","low","medium","high"]}]. AIHubMix is a multi-lab relay; exact request fields are recorded in the leading TOML comments.

The effort set matches the first-party Google entry. No budget_tokens is declared; protocol-level availability is not treated as model support.

Validation: bun validate passes on this branch. Controls were checked against the published catalog and existing model/provider entries; no inference-effect claim is made from HTTP acceptance.

[github-actions]

Action items

• [low] [possible mistake] providers/aihubmix/models/gemini-3.5-flash-lite.toml:31 - Check: Costs must be clean USD/MTok values matching the provider catalog (not float noise). Why: output = 2.499999 is almost certainly a float artifact of 2.5 / 2.50; first-party Google lists 2.50, and other AIHubMix costs use clean decimals. Action: Confirm the AIHubMix catalog price and set output to the exact published value (likely 2.5 or 2.50).

[github-actions]

No actionable findings.