feat(#2642): a github.com notification poller alongside the forge one

hive-forge-notify grows a second binary, hive-github-notify. The two
share the notification half of the job — tolerant parse, classification,
formatting, dedupe, todo delivery — and nothing else: each binary owns
its host's protocol outright.

Two binaries rather than one multi-source daemon, and rather than a
cargo feature. A feature would unify across the workspace and cost every
crate its build cache. Two binaries keep the decision in nix: forge.nix
installs the forge unit, github.nix installs the github one under
hyperhive.github.enable, so a hive built without that module has no
github poller in its closure at all — GitHub access is separable (a
tier, a policy boundary), not merely switched off. Both binaries ship
from the existing derivation, so packages.nix is untouched.

The split is real at the code level too, not just at the unit level.
source.rs is a trait; the impls live in the binaries that use them, so
neither binary links the other's protocol code and the library names no
host at all. The forge-only assigned-issue rollup moves into the forge
binary for the same reason: it asks the forge what is assigned to this
agent, which is not a notification-protocol concern.

At runtime the github unit needs a PAT at <state>/github-token, the same
dashboard-provisioned token the gh wrapper and the git credential helper
already use. No PAT: it logs why and exits 0, which is why the unit is
Restart=on-failure and not always.

Forgejo's notifications API is modelled on GitHub's, so one tolerant
parse serves both — the differences (string thread ids, PullRequest vs
Pull) are absorbed by lenient deserializers rather than a second parse
path. Thread ids normalise to String at the parse boundary; they are
only ever opaque keys. Todo keys gain a per-source prefix so the two
hosts cannot collide, and the forge's is deliberately empty to keep
existing forge todo keys stable across the deploy that lands this.

The github loop honours the server's X-Poll-Interval, re-arming only
when the server asks for a slower cadence than ours; the hint is read
before the status check, because it arrives on error and empty pages too
and that is exactly when it matters. Reading the notification stream
needs the notifications scope on the PAT, which a token minted for push
access typically lacks; the failure mode is silence, so docs/github.md
says so explicitly.

nix/agent-modules/github.nix

@@ -85,5 +85,46 @@ in
         username = x-access-token
       '';
     };
+
+    # GitHub notification poller — the github.com sibling of
+    # `hive-forge-notify` (forge.nix). Same delivery path: unread list ->
+    # per-thread summary -> todo on the harness's in-agent socket.
+    #
+    # Its own binary and its own package, so the deployment decides
+    # whether an agent gets GitHub notifications at all — the unit is
+    # installed or it isn't. That keeps the choice here rather than in a
+    # cargo feature, which would unify across the workspace and stop the
+    # two builds sharing any cached crate.
+    #
+    # The separate package is what makes that real: the per-bin
+    # extractor copies exactly one binary, so an agent that installs
+    # only the Forgejo poller has no github.com poller anywhere in its
+    # closure — not merely an unstarted unit.
+    systemd.services.hive-github-notify = lib.mkIf config.hyperhive.github.enable {
+      description = "github.com notification poller for this agent";
+      wantedBy = [ "multi-user.target" ];
+      after = [ "network.target" ];
+      environment = {
+        # Same in-agent todo socket the forge poller and the harness use
+        # (agent-service.nix) — one todo stream, two producers.
+        HIVE_AGENT_SOCKET = "/run/hive-agent/${userName}/agent.sock";
+        RUST_LOG = "info";
+        # HYPERHIVE_STATE_DIR comes from systemd.globalEnvironment; the
+        # poller reads the agent's `github-token` from under it.
+      };
+      serviceConfig = {
+        ExecStart = "${config.hyperhive.packages.hive-github-notify}/bin/hive-github-notify";
+        SyslogIdentifier = "hive-github-notify";
+        # `on-failure`, NOT `always`, for the same reason as the forge
+        # poller: this unit ships on every agent, but most agents have no
+        # PAT. The poller reports that by logging why and exiting 0, which
+        # under `always` would become a restart loop on every PAT-less
+        # agent. A crash still restarts.
+        Restart = "on-failure";
+        RestartSec = 5;
+        User = userName;
+        Group = userName;
+      };
+    };
   };
 }