Add opt-in strict checkpoint loading to from_pretrained

[YeonwooSung]

Fixes #32067

Add opt-in strict=True to from_pretrained. After loading_info is finalized, raise if there are missing or unexpected keys. Default stays False.

This matches torch.nn.Module.load_state_dict(strict=True) for those two sets. Conversion scripts can turn it on without changing For* loads of a base checkpoint.

Discussed on #32067. Arthur +1 still stood; no ACK on missing vs unexpected after 8 days, so this raises on both.

Before submitting

• Did you read the contributor guideline?
• Was this discussed/approved via a Github issue? Checkpoint validation as an option #32067
• Did you write any new necessary tests?

Who can review?

@Cyrilvallez @ArthurZucker @qgallouedec

[github-actions]

Thank you for your contribution 🤗!

CI Security Gate — automatic approval blocked

This PR was not automatically approved for CI because the security gate failed.

Possible reasons:

• The PR touches 50 or more files — only PRs with fewer than 50 changed files are automatically approved
• A changed file is outside the allowed directories (src/, tests/, docs/, utils/), has a disallowed extension (only .py, .txt, .md permitted outside tests/ and docs/), or is not .md/.yml inside docs/ — this covers files the PR deletes or renames, not only the ones it edits
• A new high-severity security issue was detected in the changed Python files (Bandit check)
• The PR touches a path this repository protects from untrusted PRs, such as the file that decides who reviews it — a maintainer must make that change in a separate PR

See the workflow run for the exact violations.

A maintainer can review and manually approve CI if a finding is a false positive.

[Cyrilvallez]

This comes from a 2 years old feature request, so not sure we actually want it haha