fix: avoid FileSchema string payload fallthrough - #5615

What does this PR address?

FileSchema.decode() accepts str inputs and starts handling them as UTF-8 file content, but the next branch was a separate if instead of an elif. As a result, string payloads still fell through to the final invalid-file-type error instead of being written to a temporary file.

This keeps PurePath inputs as paths, handles str payloads as file content, and adds a regression test for that behavior.

Validation

• .venv/bin/python -m pytest tests/unit/_bentoml_sdk/test_validators.py -q -> 2 passed in 0.42s
• .venv/bin/python -m pytest tests/unit/_bentoml_sdk -q -> 3 passed in 2.64s
• uvx ruff check src/_bentoml_sdk/validators.py tests/unit/_bentoml_sdk/test_validators.py -> All checks passed!
• uvx ruff format --check src/_bentoml_sdk/validators.py tests/unit/_bentoml_sdk/test_validators.py -> 2 files already formatted

Before submitting:

• Does the Pull Request follow Conventional Commits specification naming?
• Does the code follow BentoML code style? Targeted ruff check and ruff format --check passed for the changed files.
• Did you read through contribution guidelines and follow development guidelines?
• Did your changes require updates to the documentation? No documentation changes needed for this bug fix.
• Did you write tests to cover your changes?