How to anonymize self-citation of source code repository in IEEE double blind peer review?

I think most reviewers for IEEE Communications Letters are unlikely to visit your source code link and try and validate your code. It's fairly unusual in this field (I say this as someone in the same field.) Certainly the reviewers will not expect access to your code. So if you want to anonymize everything, that would be fine.

If you want to preserve the ability of reviewers to access your code (which is a very good thing!) and don't mind going to a little bit of extra trouble, you could always just upload an anonymized tar.gz archive containing the source code (without the version control metadata) to any file sharing site. Then use that URL in the version you submit for review. (When the paper is accepted, you can update the reference to point to your non-anonymous repository.)

Instead of source code, you should provide the executable binary, together with all the scripts to re-produce your experiments. I don't think any reviewer has the patience to read your source code.

In some conference in my field, it is suggested that the authors set up a virtual machine, e.g. a Ubuntu image, and install the compiled tool there.

In order to host these artifacts, you don't need a repository. You only need to put it in, e.g., google drive or dropbox.

Please note that if your repository is in your website, the reviewers will not visit it, for fear of being tracked.