fix(auth): specify UTF-8 encoding for auth.json read_text() calls

[JonthanaHanh]

Summary

On Windows with CJK usernames, locale.getpreferredencoding() returns GBK (or a similar multi-byte encoding). Path.read_text() without an explicit encoding argument uses the system default, so reading a UTF-8 encoded auth.json under a GBK locale raises UnicodeDecodeError, making auth completely broken for those users.

Root Cause

Three read_text() calls in hermes_cli/auth.py parse JSON files without specifying encoding="utf-8":

• _load_auth_store() (line 1115) — the main auth.json loader
• OAuth token reading (line 3704)
• Nous auth store reading (line 5139)

Meanwhile, write_text() calls and lock file operations already correctly specify encoding="utf-8".

Fix

Add encoding="utf-8" to all three read_text() calls, matching the pattern already used elsewhere in the same file.

Fixes

Fixes #69706

[teknium1]

Closing as redundant: the auth.json read sites this PR targets are already guarded on current main (hermes_cli/auth.py reads pass encoding="utf-8"), and the repo-wide bare read_text/write_text class was retired in PR #71078 (139 sites + CI linter rule). Verified per-site against main before closing. Thanks for the fix — the volume of independent PRs on this class is what escalated it to the campaign.