gen-atomicint: wrap sync/atomic types to ensure 64-bit alignment on 32-bit platforms

[AdamMagued]

Summary

On 32-bit architectures (ARM, 386, MIPS), 64-bit atomic operations (atomic.Store64, atomic.Load64, etc.) require 8-byte memory alignment. Because Int64 and Duration previously embedded raw int64 without 8-byte alignment padding or typed sync/atomic wrappers, embedding these types within larger structs on 32-bit architectures could result in a 4-byte aligned field offset, leading to:

panic: unaligned 64-bit atomic operation

Changes

• Update internal/gen-atomicint/wrapper.tmpl to wrap sync/atomic typed atomic structs (sync/atomic.Int32, sync/atomic.Int64, sync/atomic.Uint32, sync/atomic.Uint64, sync/atomic.Uintptr). In Go 1.19+, standard library sync/atomic.Int64 and sync/atomic.Uint64 embed _ align64, guaranteeing 8-byte memory alignment on 32-bit systems.
• Regenerate wrappers (int32.go, int64.go, uint32.go, uint64.go, uintptr.go, and wrapper types).
• Add 32-bit struct alignment test cases across Int64, Uint64, Float64, and Duration.
• Preserve all public methods and maintain 100% test coverage.

Fixes #105

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

Adam Magued seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}