Consolidate Randnum hex/key parsing helpers

[rdmark]

Summary

The Randnum key-file parsing logic is currently duplicated between:

• bin/afppasswd/afppasswd.c
• etc/uams/uams_randnum.c

Both files need to parse the same <afppasswd>.key format: exactly 16 hexadecimal characters representing an 8-byte DES key, with one optional trailing newline.

Recent hardening fixed both copies independently, but the duplication makes future drift likely. We should consolidate the shared parsing/validation logic into a common helper.

Background

Randnum uses a companion key file named <afppasswd>.key. The key file is read by both:

• afppasswd, when creating/updating Randnum password entries
• uams_randnum, when authenticating users or changing passwords at runtime

The logic now needs to enforce the same contract in both places:

• Accept exactly 16 hex characters
• Allow one optional trailing newline
• Reject short, long, non-hex, or otherwise malformed files
• Decode the hex into 8 bytes
• Wipe temporary buffers that may contain key material
• Avoid undefined behavior when using ctype(3) helpers

Problem

The implementation currently has two near-identical parser paths. This creates a maintenance risk:

• A validation bug may be fixed in one copy but not the other.
• Security cleanup such as explicit_bzero() on error paths must be remembered twice.
• Small C correctness details, such as casting before isdigit() / toupper(), can diverge.
• Future format changes would need duplicated edits and duplicated review.

This is especially undesirable because the parser handles secret key material.

Proposed Fix

Add a shared helper in libatalk for fixed-size hex decoding, then use it from both Randnum callers.

A general helper is preferable to exporting a Randnum-specific helper unless we want to encode the key-file newline policy there too.

Possible API shape:

int atalk_hex_decode_fixed(const unsigned char *hex,
                           size_t hexlen,
                           uint8_t *out,
                           size_t outlen);