### What happened?

Same defect as #31582, but for Claude Sonnet 5. Bedrock routes `claude-sonnet-5` through the Anthropic-compatible validator that rejects `toolSpec.strict`, so any OpenAI-format tool definition containing the `strict` field fails with a 400 from Bedrock. #31582 fixed this for Opus 4.7/4.8 (and Sonnet 4 was flagged later) by setting `bedrock_converse_supports_strict_tools: false` on the affected entries in `model_prices_and_context_window.json`, but the six Claude Sonnet 5 entries (`anthropic.claude-sonnet-5` plus the `global.`/`us.`/`eu.`/`au.`/`jp.` inference profiles) never got the flag. The gate in `bedrock_converse_supports_strict_tools()` defaults to forwarding `strict` for Anthropic models, so for Sonnet 5 the field reaches Bedrock and every tool call 400s

Expected: `strict` is dropped for Sonnet 5 the same way it is dropped for Opus 4.7/4.8 and Sonnet 4, and tool calls succeed

Impact is wider than it looks. The OpenAI Agents SDK emits `strict` in every tool definition unconditionally, so all tool calling from that SDK against Sonnet 5 on Bedrock is broken. The proxy's `/v1/responses` endpoint is broken for this model whenever tools are present even if th…