testing: add reusable MockLLMProvider for kernel-level LLMProvider trait

[Priyanshu-u07]

🔍 Issue Description

mofa-testing exports reusable mocks for ModelOrchestrator, Tool, Clock, and AgentBus but has no mock for the kernel-level LLMProvider trait defined in mofa-kernel/src/llm/provider.rs.

📌 Issue Type

• Bug
• Feature Request
• Enhancement
• Documentation
• Refactor
• Other (please specify)

📝 Description

• What is happening?
  mofa-testing (tests/src/lib.rs) provides 4 mock types:
  MockLLMBackend: mocks ModelOrchestrator
  MockTool mocks: Tool
  MockClock: mocks Clock
  MockAgentBus: mocks AgentBus

  But there is no mock for the kernel-level LLMProvider trait (chat(), chat_stream(), embedding()). Anyone
  writing agent tests at the kernel level has to create their own mock from scratch.

• What should happen instead?
  mofa-testing should export a MockLLMProvider implementing the LLMProvider trait, completing the mock
  coverage alongside the existing 4 mock types.

• Why is this needed?
  The LLMProvider trait is the primary interface agents use for LLM interaction. Without a reusable mock in the
  testing crate, test authors must write ad-hoc implementations each time. This is the only core kernel trait
  missing a mock in mofa-testing.

🎯 Proposed Solution (Optional but Encouraged)

Add MockLLMProvider to tests/src/llm_provider.rs with:

• Prompt-based response matching: map prompt substrings to preset responses (first match wins)
• Tool call simulation: return preset ToolCall responses
• Response sequences: return different responses on repeated calls (for multi-turn testing)
• Failure injection: queue errors to test error handling paths
• Request histor: record all ChatCompletionRequests for assertion
• Streaming support: chat_stream() returns deterministic chunks

Files to change:

tests/src/llm_provider.rs (new)
tests/tests/llm_provider_tests.rs (new — ~15 tests)
tests/src/lib.rs (add module + export)

📎 Additional Context

Existing MockLLMBackend has similar features (prompt matching, failure injection) but implements ModelOrchestrator, not LLMProvider

The LLMProvider trait is in crates/mofa-kernel/src/llm/provider.rs

🙋 Claiming This Issue

To avoid duplicated work:

• I'm willing to solve this issue by myself

[Priyanshu-u07]

/assign

[github-actions]

👋 @Priyanshu-u07, you've been assigned to this issue! Thank you for taking the time to contribute. Make sure to check out our contributing guidelines.