To answer your question directly: /services/FileSystem/FileServerHandler.vb

To give some friendly advice: it sounds like a good idea but ultimately it's not going to work very well.

You can't get the userid of a user unless they are logged on.  The user id can only be delivered through the dnn/asp.net authentication provider, or by lookup on the username if you have that.   You can ultimately get some users automatically if you increase the logon period so that the authentication cookie is kept for longer than default - this automatically re-logs the person on to the site *if* they checked the 'remember logon' box, and so will be available in the linkclick handler 'click'.  However, people click on emails from a variety of devices, they use different browsers, they clear their cookies, they do all sorts of things which means that they are not always going to log on automatically.  This is going to leave your data patchy and misleading, and ultimately pretty worthless.  You're also making the assumption that every user that clicks on an email is a registered user, which is faulty reasoning because any of your users could forward the email to someone else.

If you want to modify the linkclick handler I advice making your own handler and remvoving the linkclickhandler from the web.config file, and adding your own handler using the same 'linkclick.aspx' path.  This way you're not modifying the DNN Core and you leave an easier upgrade path.  You'll also need to create your own table to track the clicks, because the standard linkclick table tracks aggregate clicks, not individual clicks, and you don't want to modify the core tables.

I'm assuming your requirement is to track email open rates, and try and get an idea of which users are opening which emails.  I would suggest you look at an email service like MailChimp, because it's got more features than you could program in a years worth of work, is ready to go and will give you much higher delivery rates than just bulk emailing your users yourself.  They will give you statistics of opens on an email by email basis, which you could then re-import back into your database if you want to do analysis on it.  But the data is pretty worthless except for assessing trends in open rates - it's not particularly useful on an individual user basis because it is so unreliable, and you can't extract any meaningful information from it.