Add use_fabric_endpoint parameter to MicrosoftAzure class

[djouallah]

Summary

This PR adds support for the use_fabric_endpoint parameter to the MicrosoftAzure object store class, enabling connections to Microsoft Fabric OneLake storage.

Fixes #1356

Problem

The Python MicrosoftAzure class was missing the use_fabric_endpoint parameter that exists in the underlying Rust object_store crate. This prevented users from connecting to Microsoft Fabric OneLake storage, which requires Data Lake Storage Gen2 endpoints instead of Azure Blob Storage endpoints.

Current behavior: Defaults to https://onelake.blob.core.windows.net/...
Required behavior: Support https://onelake.dfs.fabric.microsoft.com/...

Changes

Added the use_fabric_endpoint parameter to the MicrosoftAzure class in src/store.rs:

1. Added parameter to PyO3 signature macro (line 79)
2. Added parameter to function signature as Option<bool> (line 92)
3. Added conditional builder chain call to with_use_fabric_endpoint() (lines 130-132)

This implementation follows the exact same pattern as existing boolean parameters (use_emulator, allow_http).

Usage Example

from datafusion.object_store import MicrosoftAzure
from datafusion import SessionContext
import os

# Create OneLake store with Fabric endpoint
onelake_store = MicrosoftAzure(
    container_name="tpch",
    account='onelake',
    bearer_token=os.environ["AZURE_STORAGE_TOKEN"],
    use_fabric_endpoint=True  # NEW PARAMETER
)

# Register and use
ctx = SessionContext()
ctx.register_object_store("abfss://tpch@onelake.dfs.fabric.microsoft.com/storage.Lakehouse", onelake_store, None)

Testing Note

Comprehensive end-to-end testing of this feature requires:

• An active Microsoft Fabric workspace
• OneLake storage with data
• Valid authentication tokens for OneLake access

I have manually verified that:

• The parameter is accepted without errors
• The object store can be successfully registered with a SessionContext
• Queries against OneLake storage work correctly with use_fabric_endpoint=True

The code change is minimal and follows existing patterns exactly, making it low-risk for existing functionality.

Checklist

• Follows existing code patterns
• Minimal, focused change
• Manually tested with OneLake
• No new dependencies required

[timsaucer]

Looks great, thanks for the PR. I'll wait until I see CI in the green before I merge.