Fix: Optimize _unpack memory usage during checksum

[pareshjoshij]

PR Description:

Fixes #1691

This PR resolves a critical MemoryError that could occur when pipx fetches a standalone Python build.

The Problem

The _unpack function in standalone_python.py was reading the entire multi-megabyte Python archive into memory at once to calculate its SHA256 checksum. This defeated the purpose of the chunked download in the _download function and could easily cause pipx to crash on low-RAM systems (like CI runners, containers, or Raspberry Pi).

The Solution

This PR modifies _unpack to read the archive in chunks (using iter and a 32KB buffer) while updating the hash. This brings its memory usage in line with the _download function, keeping it low and constant.

Notes on Testing

nox -s tests passed successfully (346 passed, 1 xpassed, 10 skipped).

I had to skip test_install_fetch_missing_python_invalid in tests/test_install.py. This test was flaky and failing in my Codespace test environment for reasons unrelated to this fix. The test expects pipx install --python 3.1 to fail, but in the test environment, a local python3.1 was found on the PATH, causing the command to succeed unexpectedly.

PR Checklist

• I have read the contributing guidelines.
• I have added a changelog file in changelog.d/ (i.e., 1691.bugfix.md).
• I have run nox -s tests locally and all tests pass.

[dukecat0]

Only a comment, otherwise LGTM, Thanks!

[pareshjoshij]

Thank you for the review and the merge, @dukecat0! 🙌