feat(csharp): convert cloudfetch retry from limit by # of retries to limit by time

[eric-wang-1990]

Summary

• Replace fixed 3-attempt retry with 5-minute time-budget approach using exponential backoff with jitter (matching RetryHttpHandler's pattern)
• Retry all exceptions (502 Bad Gateway from proxy, connection drops, TCP resets, HttpClient.Timeout) — catch-all, not just transport errors
• Add new adbc.databricks.cloudfetch.retry_timeout_seconds parameter (default 300s / 5 min)
• Keep adbc.databricks.cloudfetch.max_retries parameter for PowerBI backward compatibility (default -1 = not set). When set, retry loop exits if either max retries or timeout is reached first
• Add detailed retry tracing (attempt count, backoff delay, timeout budget remaining)

Problem

CloudFetch downloads used a fixed 3-retry loop with linear backoff (~1.5s total). When a proxy returns 502 or a firewall blocks cloud storage, all 3 retries exhaust within seconds — not enough time for transient issues to resolve.

Solution

Switch to a time-budget approach (same as RetryHttpHandler from #373): keep retrying with exponential backoff + jitter until the 5-minute budget is exhausted. This gives transient issues (proxy errors, firewall rules, connection drops) enough time to resolve.

	Before	After
Retry limit	3 attempts	5-minute time budget (+ optional max_retries)
Backoff	Linear (500ms, 1s, 1.5s)	Exponential with jitter (500ms → 1s → 2s → ... → 32s cap)
Total retry window	~3 seconds	Up to 5 minutes
Error classification	Retry all exceptions	Retry all exceptions (same)
max_retries param	Required (default 3)	Optional (default -1 = not set)

Test plan

• All 9 CloudFetch downloader unit tests pass
• Full build succeeds (netstandard2.0 + net8.0)
• Manual test: block cloud storage via MITM firewall, verify retries continue for 5 min
• Manual test: unblock within 5 min, verify download resumes successfully

This pull request was AI-assisted by Isaac.

[Copilot]

Pull request overview

This PR updates the C# CloudFetch downloader retry behavior from a small fixed retry count to a time-budgeted retry loop with exponential backoff + jitter, and adds a new connection parameter to configure the retry time budget.

Changes:

• Introduce adbc.databricks.cloudfetch.retry_timeout_seconds (default 300s) and wire it into CloudFetchDownloader.
• Replace the fixed retry loop with a budget-based loop that logs richer retry telemetry and supports an optional max_retries cap for backward compatibility.
• Update CloudFetch downloader tests to pass the new constructor parameters and adjust expectations/timings.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 8 comments.

File	Description
csharp/src/Reader/CloudFetch/CloudFetchDownloader.cs	Implements the new time-budget retry loop, adds retry telemetry, and plumbs through RetryTimeoutSeconds.
csharp/src/Reader/CloudFetch/CloudFetchConfiguration.cs	Adds defaults + parsing for the new retry timeout parameter and changes MaxRetries default/handling.
csharp/src/DatabricksParameters.cs	Adds the new parameter constant and updates parameter documentation for CloudFetch retry settings.
csharp/test/E2E/CloudFetch/CloudFetchDownloaderTest.cs	Updates tests to use the new retry configuration and adjusts assertions/waits for the new behavior.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[msrathore-db]

LGTM

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The thrown InvalidOperationException now includes lastException as an inner exception. Some inner exceptions (e.g., HttpRequestException) can include the original request URI in their message, which may contain pre-signed URL query parameters. Since this exception is also passed to tracing via activity?.AddException(ex, ...), the inner exception message can end up in logs/telemetry. Consider not attaching the raw exception as the inner exception, or wrapping it in a redacted exception type/message before attaching so pre-signed URL credentials can’t leak.

Suggested change

	$"Failed to download file from {sanitizedUrl} after {attemptCount} attempts over {stopwatch.Elapsed.TotalSeconds:F1}s ({retryLimits}).",
	lastException);
	$"Failed to download file from {sanitizedUrl} after {attemptCount} attempts over {stopwatch.Elapsed.TotalSeconds:F1}s ({retryLimits}).");

[eric-wang-1990]

✅ Done - Removed lastException as inner exception to prevent pre-signed URL leakage. Instead, appended the exception type name to the message: Last error: {lastException?.GetType().Name ?? "unknown"}. This gives operators enough context for diagnosis without exposing credentials.