docs: Add banner image to the top of the README

[BenWestgate]

Requested by Ben · project thread

Before: the README opens with a plain # python-codex32 heading.

After: the README opens with the illuminated python-codex32 banner ("Create, repair, and recover bitcoin backups"), the same artwork as the GitHub social preview, with the heading kept underneath it.

How: docs/images/python-codex32-banner.jpg is the full-size original the social preview was made from (2172x724), resized to 1800x600 (sharp at 2x on GitHub's README width) and saved as a progressive JPEG at quality 85 with metadata stripped: 334 KB. Nothing was cropped, recolored or repainted. The image link is absolute (raw.githubusercontent.com/.../master/...) so it also renders on PyPI, which uses README.md as the long description; the trade-off is that the rendered README on this branch shows a broken image until the file is on master.

🤖 Generated with Claude Code

https://claude.ai/code/session_01SQ3ousv9aYxskJLRZYW9JG

---

Generated by Claude Code

[BenWestgate]

build (3.13) is red here, but the failure is already on master and isn't caused by this PR.

Both build (3.13) jobs (Python package and Pylint) stop at Install dependencies. bip32 5.0.0 pins coincurve<21, and coincurve 20.0.0 ships no Python 3.13 wheel, so pip builds it from source. That build fails with Use build.verbose instead of cmake.verbose for scikit-build-core >= 0.10.

master fails the same way. On efece7d (#58), build (3.13) stops on the identical error in both Python package and Pylint. This PR only edits README.md and adds a JPEG, so it doesn't affect dependency installation. The 3.10, 3.11 and 3.12 jobs pass in Python package. The Pylint 3.10 and 3.11 jobs show as cancelled because fail-fast stopped them after 3.13 failed.

master has no fix for this yet. reviewability-v1 has no runtime dependencies (dependencies = [] after #7), so I expect the failure to go away once v1 reaches master. Porting #7 into a README PR would break the v1 order in #38, so I haven't done that. I've re-run the failed jobs once to confirm the failure is deterministic.

---

Generated by Claude Code