fix(tui): investigate log generation on Ctrl+C exit

[cyfyifanchen]

Problem

Every time the user exits Raven with Ctrl+C, Raven appears to generate or update a log. We need to determine whether this is intentional best practice, necessary TUI hygiene, or noisy behavior that should be changed.

Current context

The TUI startup path redirects loguru output to ~/.raven/logs/tui.log before launching the full-screen Ink UI. Ctrl+C is also handled in both the Python parent process and the TUI input path, depending on whether a turn is active.

Relevant questions:

• Is the log file merely opened/updated because the TUI always routes logs there?
• Is Ctrl+C writing an expected shutdown line, a warning, or an exception?
• Does a normal user exit create misleading error-looking logs?
• Should Raven suppress normal-exit noise while preserving real crash diagnostics?

Acceptance criteria

• Reproduce Ctrl+C exit and inspect the exact log lines written.
• Classify the current behavior as expected, noisy, or incorrect.
• If normal Ctrl+C writes error-looking logs, change the wording/severity or suppress it.
• If keeping the log is intentional, document why and make sure /logs or user-facing copy does not imply a crash.

[0xKT]

Investigated. On a normal exit tui.log contains only INFO lifecycle
records (RpcServer / EverosBackend start/stop) -- no ERROR lines, no
traceback. The file is created at startup by the loguru redirect (not
by Ctrl+C), which is why an exit appears to "generate" it; the Ctrl+C
path only adds a couple of closing INFO lines. So the current log
behavior is expected, not error-looking.

The one actionable, user-facing item was the TUI logs -> <path>
notice printed on every launch. #166 makes it surface only on an
abnormal child exit (and fixes the crash re-diagnosis that also fired
on SIGHUP / SIGTERM graceful exits). Reproduction was code-based -- an
interactive TTY exit cannot be reproduced headless.

Fixed by #166.