fix: reject underscores in derivation indexes

[SashaMIT]

Summary

der_to_bytes used int() on each path component. int("1_0") is 10, so m/1_0 encoded index 10. The caller already has the derivation string. An underscore is not a derivation character, and it was changing the index. Indexes now have to be digits. The existing h and ' harden suffix is unchanged.

Test plan

• Before the change, der_to_bytes("m/1_0") returned index 10 ([10, 0, 0, 0])
• tests/test_der_to_bytes.py passes after the change, including m/10 and m/1h

[netlify]

✅ Deploy Preview for specter-desktop-docs canceled.

Name	Link
🔨 Latest commit	7467e2b
🔍 Latest deploy log	https://app.netlify.com/projects/specter-desktop-docs/deploys/6ab637e28a42a80008361275

[al-munazzim]

Thanks — this is a good catch. I verified the targeted unittest locally after installing the minimal pinned Flask stack in a sparse checkout:\n\nPYTHONPATH=src python -m unittest tests/test_der_to_bytes.py\n\nResult: 3 tests passed. The change is small and correctly prevents Python's underscore digit grouping from silently changing derivation indexes.

[SashaMIT]

Thanks for taking a look, and for running the three tests.