Log Verbosity Handling in MLContext

[daholste]

When subscribing to logs from MLContext, there may not be an easy / future-proof way to know the verbosity level of each returned log statement.
Also, when subscribing to logs, for one pipeline I ran, the logger returned a total of about 255 KB of text (lots of 'Trace'-level logs). Perhaps it could be convenient for a user, when subscribing to logs, to be able to specify a minimum log severity level if desired. (For instance, a user could subscribe to 'Warning'-level logs and above.)
Thanks for reading

[eerhardt]

@CESARDELATORRE @glennc and I were just discussing this. When spitting out the Log in an ASP.NET service, it really slows down the service because of how verbose the logs are. We were thinking that

machinelearning/src/Microsoft.ML.Data/LoggingEventArgs.cs

Lines 12 to 27 in 1724da8

	public class LoggingEventArgs : EventArgs
	{
	/// <summary>
	/// Initializes a new instance of the <see cref="LoggingEventArgs"/> class.
	/// </summary>
	/// <param name="message">The message being logged.</param>
	public LoggingEventArgs(string message)
	{
	Message = message;
	}
	/// <summary>
	/// Gets the message being logged.
	/// </summary>
	public string Message { get; }
	}

Should have some sort of category/level/verbosity/etc on it.