fix: prevent verbose/debug logs from leaking to Application Insights

[qtomlinson]

Problem: Verbose and debug log messages were being sent to Application Insights even when the logger level was set to 'info'.

Root cause: The previous implementation used logger.on('data', ...) to forward Winston logs to Application Insights. This stream-based event listener operated outside Winston's transport pipeline, bypassing the transport-level filtering that enforces log level thresholds.

Solution: Replace the 'data' event listener with a custom ApplicationInsightsTransport extending winston-transport. As a proper transport, it participates in Winston's built-in level filtering, so messages below the configured threshold are never forwarded.

Additional improvements:

• Extract shared format (sanitize + errors) to logger level to avoid duplication across transports
• Use raw info.message for AI telemetry instead of printf-formatted string, since AI tracks timestamps and severity separately
• Preserve original stack trace in trackException instead of reconstructing a new Error
• Add unit tests for level filtering, error routing, and severity

Task: #739

[JamieMagee]

These loggers are never closed. Winston holds transport streams open, so Mocha can hang. Worth hoisting logger and calling logger.close() in afterEach:

Suggested change

	afterEach(() => {
	clock.restore()
	})
	afterEach(() => {
	logger.close()
	clock.restore()
	})

[qtomlinson]

The custom transport does not open a file stream or socket itself. It just forwards to trackTrace / trackException and calls the Winston callback. There is no transport-owned resource here that must be “closed” like a file transport. So the close might be optional.

There is an existing issue of the Application Insights SDK started may buffer telemetry. To avoid losing the last telemetry batch, we may want to consider calling AppInsight flush and wait for it to complete on graceful process shutdown. Can that be addressed as a separate PR?