Fix temporary service plist suffix

[oxsean]

Follow-up to #23869.

launchctl bootstrap rejects plist contents loaded from a path without a
.plist suffix with exit status 5. The temporary file introduced for
brew services run in #23869 did not specify a suffix, so macOS could not
bootstrap the generated service file.

This gives that Tempfile a .plist suffix and extends the existing test to
assert both the suffix and cleanup behaviour.

@MikeMcQuaid, I am very sorry I missed the required .plist suffix in
#23869. I reproduced the failure locally and tested this correction against
the real macOS launchctl, not only the mocked unit path:

• the same valid test plist failed with status 5 without the suffix and loaded
  successfully with the suffix;
• after applying this change to a Homebrew checkout on current main,
  brew services run cc-connect succeeded;
• launchctl print showed the generated .plist path and all values from
  ~/.homebrew/services/cc-connect.env;
• the temporary plist was removed after bootstrap while the service remained
  running.

Steps to reproduce the regression on macOS before this change:

mkdir -p "${HOMEBREW_USER_CONFIG_HOME:-$HOME/.homebrew}/services"
echo 'CC_LOG_MAX_BACKUPS=3' > "${HOMEBREW_USER_CONFIG_HOME:-$HOME/.homebrew}/services/cc-connect.env"
brew services run cc-connect

The last command fails while executing launchctl bootstrap against a
temporary path without a .plist suffix.

Verification:

./bin/brew tests --only=services/cli:1228
./bin/brew lgtm --online

---

• Have you followed our Contributing guidelines?
• Have you checked for other open Pull Requests for the same change?
• Have you explained what your changes do? Performance claims (e.g. "this is faster") must include brew benchmark results.
• Have you explained why you'd like these changes included, not just what they do?
• For bug fixes, have you given step-by-step brew commands to reproduce the bug?
• Have you written new tests (excluding integration tests)? Here's an example.
• Have you successfully run brew lgtm (style, typechecking and tests) locally?

---

• I did not use AI/LLM to create this PR, or I disclosed the tool/model below and reviewed its output; I did not attribute commits to AI and will answer maintainer questions and review comments myself without AI/LLM.

OpenAI Codex (GPT-5) was used to prepare and test this follow-up under the
contributor's direction. The diff was reviewed, the regression test was run
red then green, the full brew lgtm --online gate passed and the change was
verified with a live macOS launchctl service run as described above.

---

[MikeMcQuaid]

Thanks!