fix: use utf-8 in CI smoke server script

[Ghraven]

Problem

scripts/ci_smoke_server.py writes and reads text artifacts without specifying an encoding. On Windows or other non-UTF-8 default locales, the generated app script or captured server log can become locale-dependent even though the script content is text.

Before / after

Before:

• The smoke server log was opened with the platform default text encoding.
• The failure log dump used Path.read_text() without an explicit encoding.
• The generated Burr app script used Path.write_text() without an explicit encoding.

After:

• All three text boundaries use encoding="utf-8".
• Binary/network handling is unchanged.

Verification

• python -m py_compile scripts/ci_smoke_server.py

[Ghraven]

Thank you for the review and merge, I appreciate it!