Think in terms of at least thousands of years to brute force a ssh key. ssh probes are short lived, directed against weak security, have no intelligence behind them and stand no serious chance of succeeding.
While I agree that SSH should rely on strong passwords or keys, I still lean towards giving an attacker as little information as possible; if they don't know your username then it's much harder to try to break in, which is why login programs no longer tell you that you entered an invalid user ID.
I don't know whether it's still the case, but a lot of network services used to have the host name in the banner, which would allow remote access to the name.
User ID known and a 2048 bit key to brute force. What is our expectation that this will be done within 1,000,000 years? How about zero? So, a 100% secure account.
User ID unknown so the attacker guesses 100, none of which exist on that machine. Expectation of success? Zero. So again we have a 100% secure account. But not knowing the username hasn't increased security. I'm prepared to give out a list of user IDs used here and reckon it will not detract from the security of ssh communication.