fix(providers/openai): fix Responses replay continuity

[ibetitsmike]

Worked on by Mux on Mike's behalf.

Summary

• Replay stored reasoning items as item_reference when store=true and a persisted reasoning item ID is available.
• Replay provider-executed web search item references only when they immediately follow a replayed reasoning reference, otherwise skip them instead of sending a bare ws_* item.
• Validate explicit Responses replay before send so local function_call items pair with exactly one later function_call_output, and ws_* item references have a preceding rs_* reference.
• Add contract tests for reasoning plus web search replay, missing reasoning metadata, store=false replay, and malformed function-call ordering.

Tests

• go test ./providers/openai -run 'TestResponsesToPrompt_(ReasoningWithStore|ReasoningWithWebSearchCombined|WebSearchRequiresReasoningReference|ReasoningWithFunctionCallCombined|WebSearchProviderExecutedToolResults)|TestPrepareParams_(SkipsProviderExecutedToolReferences|ValidatesFunctionCallOutputPairing)|TestValidateResponsesInput_WebSearchReferenceRequiresReasoning' -count=1
• go test ./providers/openai -count=1
• git diff --check

[ibetitsmike]

/coder-agents-review