AArch64: Fix register access for LSE atomics

[oleavr]

Your checklist for this pull request

• I've documented or updated the documentation of every API function and struct this PR changes.
• I've added tests that prove my fix is effective or that my feature works (if possible)

Detailed description

The AArch64 LSE atomics (CAS, CASP, SWP, and the LD/ST families) had empty operand-access
entries, so cs_regs_access() reported neither the value registers read nor the destination written. E.g.
swp w0, w1, [x2] reported only x2 read, nothing written.

Fixes:

1. Populate the operand-access table in AArch64MappingInsnOp.inc.
2. Handle CASP register pairs in AArch64_reg_access() — the printer leaves the second pair element
   without access info. Operands 0–1 are read+written, 2–3 read.

No API/struct changes.

Test plan

suite/regress/test_arm64_atomics.py asserts the read/written sets from regs_access() across the affected
families. Fails before the fix, passes after.

python3 suite/regress/test_arm64_atomics.py

Closing issues

None.

[oleavr]

The base was briefly wrong when I opened this, so the labeler applied every arch/binding label. Base is now v5; only AArch64 and Python apply. I can't edit labels myself — could a maintainer clear the rest? Thanks!

[Rot127]

Thanks! See the comment.

[Rot127]

This needs to be added to the bindings/python/Makefile on the v5 branch. Otherwise it is not run in the CI.

[oleavr]

Done — added a REGRESS_TESTS entry to the check target so CI runs it.

Note: the existing loop ran ./$t from bindings/python/, but the tests live in tests/, so it was already failing before any regress test. Fixed it to cd tests first. Can split that out if you'd rather.

[Rot127]

Let me merge this one first: #2954

[Rot127]

Freaking v5 testing. I am so close to rewrite it.

Anyways, the makefile way is broken and I removed that step from the workflow.

Can you please revert the Makefile changes and add the test to ./bindings/python/tests/test_all.py please?
Then rebase and we can merge.

Sorry for all the back and forth. I haven't looked at the v5 testing for a long time and forgot where is what done.