macro/align: Use ALIGN_UP and ALIGN_DOWN uniformly

[W-M-R]

Note: Please adhere to Contributing Guidelines.

Summary

macro/align: Use ALIGN_UP and ALIGN_DOWN uniformly

Impact

Update this section, where applicable, on how change affects users,
build process, hardware, documentation, security, compatibility, etc.

Testing

Update this section with details on how did you verify the change,
what Host was used for build (OS, CPU, compiler, ..), what Target was
used for verification (arch, board:config, ..), etc. Providing build
and runtime logs from before and after change is highly appreciated.

[xiaoxiang781216]

please fix ci error:

D:\a\nuttx\nuttx\sources\nuttx\mm\iob\iob_initialize.c(63,29): error C2057: expected constant expression [D:\a\nuttx\nuttx\sources\nuttx\vs2022\mm\mm.vcxproj]
  iob_trimhead.c
D:\a\nuttx\nuttx\sources\nuttx\mm\iob\iob_initialize.c(63,29): error C2466: cannot allocate an array of constant size 0 [D:\a\nuttx\nuttx\sources\nuttx\vs2022\mm\mm.vcxproj]
D:\a\nuttx\nuttx\sources\nuttx\mm\iob\iob_initialize.c(63,16): error C2133: 'g_iob_buffer': unknown size [D:\a\nuttx\nuttx\sources\nuttx\vs2022\mm\mm.vcxproj]

[yamt]

it isn't obvious HISTFRACTION * sizeof(HISTCOUNTER) is a power of two.

is it possible to have a power-of-two assertion in ALIGN_UP itself?
this PR as it is seems a bit fragile to me.

does ALIGN_UP actually produce a better code than ROUNDUP with a power of two y with modern compilers? otherwise, maybe it's safer to drop the power-of-two assumption from ALIGN_UP.

[W-M-R]

I think what you said makes sense. Let me open a new PR to fix this problem.

[W-M-R]

Fix it in the patch https://github.com/apache/nuttx/pull/15493/files. Sorry, I didn't notice this problem before.