fix: pre-existing issues in server.py and common.py (raise_if_dead, has_minimum_version)

[tony]

Summary

Three pre-existing issues surfaced during a fresh code review of PR #636. This PR addresses them as atomic commits against master.

• P1common(fix[has_minimum_version]): has_minimum_version was calling get_version() twice, spawning two subprocesses for no reason. Now cached in a local variable.
• P2server(docs[raise_if_dead]): raise_if_dead had no Raises section in its docstring. Both TmuxCommandNotFound (missing binary) and CalledProcessError (server not running) are now documented.
• P3server(fix[raise_if_dead]): subprocess.check_call in raise_if_dead inherited the parent process's stdout/stderr, leaking list-sessions output to the terminal. Now redirected to subprocess.DEVNULL.

Test plan

• uv run ruff check . --fix --show-fixes — clean
• uv run ruff format . — no changes
• uv run mypy — no issues (60 source files)
• uv run py.test -x -q — 885 passed, 1 skipped

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 52.44%. Comparing base (d8b7127) to head (218c0fb).

Additional details and impacted files

@@ Coverage Diff @@## master #641 +/- ##
=======================================
Coverage 52.44% 52.44% =======================================
Files 26 26 Lines 3726 3726 Branches 747 747 =======================================
Hits 1954 1954 Misses 1468 1468 Partials 304 304 

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[tony]

Flagging that this PR is more current than its title suggests, and that it silently fixes an issue filed four months after it.

The two things the title and body advertise have already landed on master independently — has_minimum_version caching (ea561463) and the Raises docstring (85ace722). What remains is a single commit touching one file, and its whole diff is:

-subprocess.check_call([resolved, *cmd_args])
+subprocess.check_call(
+ [resolved, *cmd_args],
+stdout=subprocess.DEVNULL,
+stderr=subprocess.DEVNULL,
+ )

That is exactly the fix #711 asks for: Server.raise_if_dead() shells out unredirected, so tmux's no server running on … / error connecting to … lands on the caller's fd 2 and cannot be intercepted with contextlib.redirect_stderr. Server.is_alive() goes through self.cmd() and is quiet, so the two liveness primitives behave inconsistently.

Worth retitling this PR to name that fix, and deciding one open question here rather than in a second PR: DEVNULL stops the leak but discards tmux's diagnostic entirely. Routing raise_if_dead() through self.cmd() would keep the stderr on the returned object where a caller can read it — but it would stop raising subprocess.CalledProcessError, which tests currently pin.