Bugfix: unaligned access in MIPS getdents64

[ljucovic]

Checklist

Which kind of PR do you create?

• This PR only contains minor fixes.
• This PR contains major feature update.
• This PR introduces a new function/api for Qiling Framework.

Coding convention?

• The new code conforms to Qiling Framework naming convention.
• The imports are arranged properly.
• Essential comments are added.
• The reference of the new code is pointed out.

Extra tests?

• No extra tests are needed for this PR.
• I have added enough tests for this PR.
• Tests will be added after some discussion and review.

Changelog?

• This PR doesn't need to update Changelog.
• Changelog will be updated after some proper review.
• Changelog has been updated in my PR.

Target branch?

• The target branch is dev branch.

One last thing

• I have read the contribution guide

---

[retrocpugeek]

Cross-linking: I hit this same alignment bug on a big-endian MIPS guest and opened #1635 (issue + repro) and #1636 (fix + regression tests).

Brief summary of #1636's fix — it covers both getdents and getdents64:

• Rounds each record up with (d_reclen + (n - 1)) & ~(n - 1) (canonical round-up, vs (d_reclen + n) which over-pads).
• For the legacy getdents record it places the alignment padding between d_name and d_type, keeping d_type in the last byte (offset d_reclen - 1). Padding the legacy record without relocating d_type moves the consumer's read into the pad bytes — I suspect that's what got Bugfix: unaligned access in MIPS getdents64 #1419 reverted.

Supersedes #1425. Not trying to step on this PR — happy to consolidate whichever direction the maintainers prefer.