Improve test coverage of the agent service

[bobbai00]

Task Summary

The agent service (agent-service/) currently has low automated test coverage. A bun test --coverage run reports ~51.6% line / 54.2% function coverage overall, and the gaps are concentrated in the core logic rather than helpers:

Module	% Lines	Role
util/context-utils.ts	1%	builds per-step LLM context
tools/workflow-execution-tools.ts	5%	operator execution tool
api/workflow-api.ts	8%	workflow CRUD client
api/compile-api.ts	10%	compile client
util/workflow-utils.ts	10%	workflow helpers
agent/prompts.ts	15%	system prompt builder
tools/workflow-crud-tools.ts	22%	add/modify/delete tools
agent/texera-agent.ts	32%	core ReAct loop (sendMessage)
server.ts	50%	REST routes + websocket

The well-covered files (result-formatting, tools-utility, auto-layout, workflow-result-state) are mostly leaf utilities, so the headline number overstates how much of the agent's behavior is actually exercised.

Add unit/integration tests for, at minimum:

• the texera-agent.ts ReAct loop — step-tree construction, HEAD/checkout, stop/abort, error steps, settings effects;
• the CRUD and execution tools (positive + error/[ERROR] paths);
• context assembly in context-utils.ts;
• the API clients (workflow-api, compile-api, backend-api) against mocked HTTP, including bad/malformed responses.

Tests should cover both positive and negative paths (invalid input, missing config, backend errors), per the project testing guidelines.

Task Type

• Refactor / Cleanup
• DevOps / Deployment / CI
• Testing / QA
• Documentation
• Performance
• Other