Derive VanguardMode from VanguardParams and has_onion_svc

From #tor-dev

2024-04-11 11:24:25	 *	gabi took a detour from vanguard persistence to hack on parts of arti#1272. I feel like we don't actually need a vanguards config at this point. ISTM all the configuration comes from the NetParameters (which can be overriden in the config)
2024-04-11 11:24:26	--	Notice(tor): https://gitlab.torproject.org/tpo/core/arti/-/issues/1272 - Add vanguard configuration options
2024-04-11 12:15:46	+nickm	gabi: I think you're likely right wrt configuration.
2024-04-11 12:16:24	gabi	ack. I think I'll remove the VanguardConfig for now (right now it's just an empty struct)
2024-04-11 12:16:25	+nickm	gabi: Sometimes we might want to provide a more ergonomic way to override consensus parameters, but I don't think this is one of those times
2024-04-11 12:16:32	gabi	right

So this ticket is about

• removing the VanguardConfig and deriving the VanguardMode from the VanguardParams like so

    /// Get the current [`VanguardMode`].
    ///
    /// If we are not running an onion service, we use the `vanguards_enabled` mode.
    ///
    /// If we *are* running an onion service, we use whichever of `vanguards_hs_service`
    /// and `vanguards_enabled` is higher for all our onion service circuits.
    fn mode(&self) -> VanguardMode {
        if self.has_onion_svc {
            std::cmp::max(
                self.params.vanguards_enabled(),
                self.params.vanguards_hs_service(),
            )
        } else {
            self.params.vanguards_enabled()
        }
    }

  where has_onion_svc is a boolean value specifying whether arti is running as an onion service or not. For context, see torspec!258 (comment 3011734)
• setting has_onion_svc in the VanguardMgr to true or false, depending on whether the config enables an onion service or not (this also needs to be handled in reconfigure()).

This will likely involve rethinking our reconfigure() APIs. Currently, reconfigure() takes a &TorClientConfig, which doesn't contain the onion svc config, so when the VanguardMgr is reconfigure()d in response to a config change, we have no way of telling if its has_onion_svc flag needs to be updated.

Moreover, in !2035 (merged), the vanguards API was designed to purge all HS circuits on vanguard "mode" changes. However, that design assumed the vanguard mode (full/lite/disabled) would be read from the config, which will no longer be true after this ticket is implemented (the vanguard mode will be derived from the consensus params and the onion svc config). The VanguardMgr will need to be able to notify the HsCircPool that it needs to retire all circuits whenever the vanguard mode changes (which can be due to a config change or a consensus change).