test(connector-opencode): assert the event channel is core's derivation by identity, not one sampled value

[davidfarah2003]

What

The events-arm suite for the OpenCode connector asserted that the connector's
declared event channel matched core's derivation for one sample principal, by value:

// principal = { owner: "local", actor: "ollie" }
opencodeConnector.eventChannel?.(principal) === eventChannel(principal)

That is the right comparison at the wrong strength. A re-implementation of the
derivation that returned the correct string for ollie and diverged elsewhere would
pass the cell unchanged. What makes the property true in the shipped connector is not
the sampled value: it is that the connector's field IS core's function rather than a
copy, so the channel the manager mints the grant for and the subject the session
publishes to are decided in exactly one place.

This adds the function-identity assertion:

opencodeConnector.eventChannel === eventChannel

mirroring the assertion the codex events-arm suite already carries. The existing
by-value cell stays, because it additionally proves the derivation is keyed on the
principal rather than the display name.

Test-only. The property is already true; no behaviour changes. Closes #601.

Proof

• Baseline: smoke:opencode-events-arm reports 16 passed, 0 failed.
• Mutation, cloning the assignment to eventChannel: (p) => eventChannel(p): KILLED,
  red on the named assertion "not a re-implementation"; mutation-proof reports the
  suite discriminates.
• The gap shown directly: under that same clone, exactly one cell fails, the new
  identity cell, and the by-value cell is among the 15 that pass. The by-value cell
  alone could not have caught the clone; the new cell does.
• pnpm typecheck exits 0.

Sibling gap, filed separately

The claude-code events-arm suite has the identical weakness (it asserts the declared
channel by value on one principal and no cell asserts function identity). Rather than
widen this PR past the issue it closes, that is filed as #713 so the weakness is on the
record with its reason, which is the disposition #601 itself argued for.