CI: upload server binary

[samcowger]

WIP

[samcowger]

Hmm. With this change, I'd hoped to be able to have prospective users simply download a binary of the server, put it on their $PATH, and be set. However, it seems like Cerberus and CN, at build time, populate their build environment with some files, and then rely on being able to find them there (or in some alternate locations - see Cerberus's util/cerb_runtime.ml for the search logic) at runtime. So, exfiltrating the server binary alone isn't enough - it needs to have some sort of access to those runtime files.

I see a few possible options to work around this:

• Upload those files alongside a server binary, and have instructions for how to point the server at them.
• Change Cerberus to compile those files into itself rather than searching for them on disk at runtime.
• Enhance the server build process so that it, rather than Cerberus, carries those files' contents in its binary, serializes them to disk at runtime, and points Cerberus to where it serialized them.

The first seems easiest, but also more fiddly than necessary from a user perspective. The second might be a bit of an invasive change to Cerberus. The third probably offers the cleanest user experience, but would probably result in a pretty messy build process for the server.

I'll put this aside for now and ponder how/whether to move forward.