FIX: read initializer scripts as UTF-8 in add-initializer

[MinYi Xie (ppcvote)]

Description

add-initializer reads the local script without an encoding, so on Windows it is decoded with the ANSI code page instead of UTF-8: pyrit/cli/pyrit_scan.py:822 (aiofiles.open(script_path)) and pyrit/cli/pyrit_shell.py:399 (script_path.read_text()), both at b0dba3e. Under -X warn_default_encoding both reads report 'encoding' argument not specified.

Any initializer holding a non-ASCII character is affected wherever the OS default text encoding isn't UTF-8, which includes Windows. On a cp950 machine (Python 3.11.6), add-initializer on a Spanish initializer prints Registered initializer 'spanish_init' and returns 0, while the text passed to register_initializer_async is "Eres un asistente 繳til..." instead of "Eres un asistente útil...". Feeding that text to CustomInitializerStorage.save_script and reading it back keeps the damage, so the prompt the initializer sets changes while the CLI reports success. A file with Japanese text fails instead, with 'cp950' codec can't decode byte 0xe6 in position 76: illegal multibyte sequence, and so does a renamed copy of PyRIT's own pyrit/setup/initializers/targets.py, which contains an em dash. Read as cp1252, the same bytes give café, año, niño with no error at all.

Python source is UTF-8 by default (PEP 3120), and the rest of this path already agrees: pyrit/cli/_config_reader.py:68, pyrit/registry/custom_initializer_storage.py:70,80 and pyrit/registry/components/initializer_registry.py:485 all pass encoding="utf-8". The fix passes encoding="utf-8" at both reads.

Tests and Documentation

• New test_handle_add_initializer_reads_source_as_utf8 (scan) and test_success_path_reads_source_as_utf8 (shell) register a UTF-8 file with accented Latin and Japanese text and assert the uploaded content matches the file.
• New test_reads_script_with_explicit_utf8_encoding and the existing test_handle_add_initializer_reads_with_aiofiles assert the encoding argument, so they fail on any OS. That existing test pinned open(path) and is updated to open(path, encoding="utf-8").
• The current tests miss this: test_main_add_initializer_success writes ASCII content, and test_handle_add_initializer_reads_with_aiofiles mocks aiofiles.open.
• Reverting only the source change fails all four tests on this machine; with it tests/unit/cli is 242 passed. ruff 0.16.6 check and format --check pass on the changed files, and ty 0.0.78 passes on the two modules.
• No doc change: no encoding is documented for initializer scripts. I did not open an issue first, happy to file one if you prefer.