chore: normalize line endings on two validation specs

token-validation.service.spec.ts and state-validation.service.spec.ts
were checked in with mixed CRLF/CR/LF line endings, which git's
auto-detection treated as binary (`-text`). That made every previous
PR touching them produce a wall-of-noise diff because git couldn't
do line-level comparison.

Re-encode to LF to match the rest of the repo (core.autocrlf = input).
No semantic changes — only line-ending bytes differ.

This is committed separately so the actual test conversion in the
next commit shows a clean diff.