fix(google-pricing): rectify audio input, Veo, and Imagen pricing entries

[akhilmmenon]

PR Description

Summary

• Updates Google pricing data in models/pricing/google.json for the pricing/google-update branch.
• Corrects audio-input related pricing coverage and aligns Veo pricing semantics.
• Fixes Imagen pricing representation to match the repository’s existing pricing unit convention.

Why

• Ensures Google pricing entries are consistent with provider documentation and internal pricing conventions.
• Prevents billing discrepancies caused by incomplete modality coverage or inconsistent unit representation.

Test Plan

• Validate changed model rows in models/pricing/google.json against Google pricing docs.
• Run repository checks/tests used for pricing data validation.
• Spot-check affected families (audio-input models, Veo models, Imagen models) for expected rates and units.

[Copilot]

Pull request overview

Updates Google provider pricing configuration to align token/audio/cache dimensions and video/image pricing semantics with current documentation and repository conventions.

Changes:

• Expands Google Gemini entries with audio-input pricing, cache read/write rates, and adds/adjusts tool-unit pricing.
• Refactors Veo pricing units away from “default_*” placeholders into explicit duration/resolution units and adds missing structural blocks (batch_config, additional_units).
• Adjusts Imagen/Gemini image-preview pricing representation and adds several new model entries (e.g., live preview, native audio preview, Lyria, Veo lite).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

fixed_cost is added under pricing_config for the new Lyria models, but fixed_cost is not part of the documented PricingConfig schema (openapi.yaml doesn’t define it). If fixed_cost is a supported pricing dimension, the schema (and any validators/consumers) should be updated to include it; otherwise these entries may be silently dropped or treated as invalid by typed clients.

[Copilot]

batch_config includes request_audio_token, but the documented BatchConfig schema only includes request_token, response_token, and cache_read_input_token (see openapi.yaml → BatchConfig). If batch audio pricing is intended, the schema and any downstream cost calculation/validation should be updated to support it; otherwise omit request_audio_token from batch_config so clients don’t ignore or reject the field.

[Copilot]

This PR introduces a new additional_units.maps price, but the PR description only calls out audio-input, Veo, and Imagen adjustments and the repo docs don’t currently list maps as a known additional unit. Please either (a) update the PR description (and ideally README/openapi docs) to cover the new unit and confirm downstream consumers recognize it, or (b) remove/rename it to an existing documented unit if it was meant to be search/web_search.