Fix/hide reviewer prompt from terminal

[Rahul-2006]

Summary

Fixes #2410 by preventing reviewer prompts from being reprinted in the terminal.

• Updated the Claude Code, Codex, and OpenCode reviewer adapters so ReviewMessage() returns an empty string instead of the full prompt.
• Updated tests to verify the new behavior.

Validation

• All reviewer adapter tests pass: go test -v ./internal/adapters/reviewer/...
• Reviewed with CodeRabbit, no additional issues were reported.

@harshitsinghbhandari can you please review the pr and suggest any changes.

[neversettle17-101]

This should not drop the re-review instruction entirely.

The requirement is: reviewer prompts should not be shown/reprinted to the user in the terminal, similar to how orchestrator/worker system prompts are not exposed. But the reviewer must still receive meaningful task context for every subsequent review pass/new commit.

Right now this PR makes ReviewMessage() return an empty string. In the current launcher flow, Notify() calls ReviewMessage() and sends that string to the already-running reviewer pane. If the message is empty, the reviewer gets no new PR/run id/review queue context for the next pass, so this can silently break re-review behavior.

Please change the approach so the prompt is hidden from user-facing output but still delivered to the reviewer through a non-displayed mechanism, or otherwise send a minimal visible trigger that lets the reviewer load the queued task context. The fix should include a test that proves an already-running reviewer can be notified of a subsequent review task without reprinting the full prompt.