[Bug] generate_csv_file drops keys missing from the first row

Description

FileGenerationTools.generate_csv_file takes CSV headers from the first dict only. Later rows with extra keys silently drop those fields.

Steps to Reproduce

from agno.tools.file import FileGenerationTools

tools = FileGenerationTools()
result = tools.generate_csv_file([
    {"name": "Ada"},
    {"name": "Ben", "city": "London"},
])
print(result.files[0].content.decode("utf-8"))

Expected Behavior

The CSV includes every key that appears in any row. Missing values are empty cells.

Actual Behavior

The city column is omitted. Ben's city is lost.

Environment

• OS: Windows 11
• Agno Version: current main
• Additional Environment Details: Python 3.13

Possible Solutions (optional)

Union keys across all dict rows, preserving first-seen order.