Fix encoding bug: replace Encoding.ASCII with Encoding.Latin1

[carlbennett]

Summary

• Battle.net protocol specifies ISO 8859-1 (Latin1) for string fields
• Encoding.ASCII silently corrupts bytes above 0x7F to ? before hashing, causing auth failures for extended characters
• Fixed in NLS, OldAuth, DataBuffer, DataReader, and all protocol string helpers

Test plan

• Auth with usernames/passwords containing extended Latin1 characters
• Protocol string fields round-trip correctly

🤖 Generated with Claude Code

[carlbennett]

Verified: usernames with extended Latin1 characters exist in the wild. @islanti maintains a cache of Battle.net usernames (illys_list.html) collected from USEast, USWest, Europe, and Asia realms, and scanning it confirms bytes above 0x7F appear in real accounts. Encoding.ASCII would silently corrupt those to ? before hashing, producing a wrong hash and failing auth. This fix is necessary for correct behavior.

$ cat illys_list.html | od -An -tu1 | tr ' ' '\n' | grep -v '^$' | sort -n | uniq -c | sort -rn | awk '{printf "dec=%3d (0x%02X): %d\n", $2, $2, $1}' | grep -v '0x[0-6].' | grep -v '0x7[0-F]'
dec=198 (0xC6): 1
dec=179 (0xB3): 1
dec=154 (0x9A): 1
dec=128 (0x80): 1