Fix multipart upload parsing edge cases

[freelamb]

Motivation

• Harden multipart POST handling to correctly parse variable part headers and avoid assumptions about optional headers.
• Prevent truncation when uploaded content contains boundary-like text and ensure interrupted or malformed uploads do not leave partial files on disk.
• Add targeted unit tests for upload parsing regressions and record a bugfix release.

Description

• Replace the brittle header-reading logic in deal_post_data() with a loop that reads part headers until the blank separator and extracts the filename from combined headers, using helper functions read_upload_line() and is_boundary_line() to robustly detect boundaries.
• Add validation to reject negative Content-Length values and ensure the upload target from translate_path() is an existing directory before writing the file.
• Preserve uploaded content that omits an optional part Content-Type header and avoid treating content that merely contains boundary-like text as a delimiter by matching only full boundary delimiter lines.
• Ensure partial files are removed on failure by deleting the incomplete target_path when the upload does not complete successfully, and bump __version__ to 0.3.4 and update CHANGELOG.md accordingly.
• Add new tests in tests/test_simple_http_server.py (UploadParsingTests) and a DummyUploadHandler to cover: missing part Content-Type, content containing boundary-like text, and interrupted uploads leaving no partial file.

Testing

• Ran python3 -m py_compile simple_http_server.py successfully.
• Ran python3 -m unittest discover -s tests and all tests passed (8 tests ran, OK).
• Attempted flake8 checks but flake8 is not installed in the environment, so those checks were not executed.

---

Codex Task

[freelamb]

issues