|WARNING| [v2] common/mlx5: fix mkey ordering enum

[dpdklab at iol.unh.edu]

Test-Label: ai-code-review
Test-Status: WARNING
https://dpdk.org/patch/171829

_AI Code Review_

=== Patch Review: series.patch (via Claude) ===
Review date: 2026-10-22

# DPDK Patch Review

## ERRORS

**1. Enum value removal without ABI versioning**

Line removed:
```c
MLX5_MKC_RAW_ORDERING_SAO = 0x1,
```

This removes `MLX5_MKC_RAW_ORDERING_SAO` from a public enum.
If this enum is part of the installed API (in `drivers/common/mlx5/mlx5_prm.h`),
removing a value is an ABI break.
Existing compiled code that references `MLX5_MKC_RAW_ORDERING_SAO`
will fail to link or will silently use the wrong value (0x1 now maps to `MLX5_MKC_RAW_ORDERING_RO`).

**Fix:** Check if `mlx5_prm.h` is in the driver's `headers` in meson.build.
If it is installed, this change requires ABI versioning:
deprecate `MLX5_MKC_RAW_ORDERING_SAO` in this release with a comment,
and remove it in the next ABI-breaking window.
If the header is driver-internal only (not installed), this is acceptable
but verify no remaining uses of `MLX5_MKC_RAW_ORDERING_SAO` exist in the tree.

---

## Evidence Check

- The removed line `MLX5_MKC_RAW_ORDERING_SAO = 0x1,` is visible in the diff as a deletion.
- The enum value collision: the old `RO = 0x2` becomes `RO = 0x1`, occupying the slot previously held by `SAO`.
- ABI impact depends on whether the header is installed; the finding stands subject to that check.

---

Review-Result: ERROR