fix: populate inputSchema and outputSchema for coded agents

[mjnovice]

Summary

Fixes the extraction of inputSchema and outputSchema when reporting eval runs to Studio Web, ensuring proper schema population for both coded and low-code agents.

Changes

• Enhanced _extract_agent_snapshot() to support low-code agents by reading schemas directly from agent JSON files (e.g., agent.json)
• Maintained existing behavior for coded agents that read from entry-points.json
• Added comprehensive logging throughout the progress reporter for better debugging of eval run reporting
• Improved error handling for missing or invalid entrypoints

Technical Details

The method now follows this logic:

1. If entrypoint is a JSON file (low-code agent), read inputSchema and outputSchema from the agent file
2. Otherwise, fall back to reading from entry-points.json for coded agents
3. Return empty schemas if neither source is available

Testing

Added comprehensive test suite for agent snapshot extraction covering:

• Reading schemas from entry-points.json for coded agents
• Handling missing files and entrypoints
• Proper warning when entrypoint is None or empty
• Correct camelCase serialization for API payloads

Development Package

• Use uipath pack --nolock to get the latest dev build from this PR (requires version range).
• Add this package as a dependency in your pyproject.toml:

[project]
dependencies = [
  # Exact version:
  "uipath==2.2.38.dev1010423494",

  # Any version from PR
  "uipath>=2.2.38.dev1010420000,<2.2.38.dev1010430000"
]

[[tool.uv.index]]
name = "testpypi"
url = "https://test.pypi.org/simple/"
publish-url = "https://test.pypi.org/legacy/"
explicit = true

[tool.uv.sources]
uipath = { index = "testpypi" }

[tool.uv]
override-dependencies = [
    "uipath>=2.2.38.dev1010420000,<2.2.38.dev1010430000",
]