fix(csharp): apply CloudFetch timeout and remove redundant HttpClient

[eric-wang-1990]

What's Changed

This PR addresses two issues in the CloudFetch implementation and one fix in the proxy test infrastructure:

1. Apply CloudFetch Timeout Configuration

Problem: The adbc.databricks.cloudfetch.timeout_minutes configuration parameter was being read but never applied to the HttpClient. The HttpClient used its default 100-second timeout instead of the configured value (default: 5 minutes).

Solution: Added timeout configuration in CloudFetchDownloader constructor:

_httpClient.Timeout = TimeSpan.FromMinutes(config.TimeoutMinutes);

Impact: The CloudFetch HTTP client now correctly respects the user-configured timeout value, allowing for proper timeout behavior in CloudFetch downloads.

Files Changed:

• csharp/src/Reader/CloudFetch/CloudFetchDownloader.cs

2. Remove Redundant HttpClient from CloudFetchDownloadManager

Problem: CloudFetchDownloadManager accepted an HttpClient parameter but never used it for any operations. It only disposed of it, which is incorrect since it doesn't own the resource. The actual HTTP operations are performed by CloudFetchDownloader.

Solution: Removed the redundant httpClient parameter from:

• CloudFetchDownloadManager constructor and field
• CloudFetchReaderFactory.CreateThriftReader() call
• CloudFetchReaderFactory.CreateStatementExecutionReader() call

Impact: Cleaner code architecture with proper resource ownership. The HttpClient is now owned solely by CloudFetchDownloader, which actually uses it for HTTP operations.

Files Changed:

• csharp/src/Reader/CloudFetch/CloudFetchDownloadManager.cs
• csharp/src/Reader/CloudFetch/CloudFetchReaderFactory.cs

3. Fix Proxy Timing for Test Reliability

Problem: The mitmproxy addon was disabling failure scenarios AFTER the delay completed, causing race conditions. When the client timed out and retried, the scenario was still enabled, triggering the delay again.

Solution: Modified mitmproxy_addon.py to disable the scenario immediately when triggered, before the delay starts:

self._disable_scenario(scenario_name)  # Disable BEFORE delay
await asyncio.sleep(duration_seconds)

Impact: CloudFetch timeout tests now behave correctly - only the first request is delayed, and retry attempts succeed immediately.

Files Changed:

• test-infrastructure/proxy-server/mitmproxy_addon.py

Testing

All 5 CloudFetch proxy tests now pass:

• ✅ CloudFetchConnectionReset_RetriesWithExponentialBackoff
• ✅ CloudFetchExpiredLink_RefreshesLinkViaFetchResults
• ✅ NormalCloudFetch_SucceedsWithoutFailureScenarios
• ✅ CloudFetchTimeout_RetriesWithExponentialBackoff
• ✅ CloudFetch403_RefreshesLinkViaFetchResults