Fix misalignment of small structs in a Vector (C++)

[bexcite]

Fixes the bug that was introduced by: #7520

Closes: #8024

The issue:
Small structs that are less than uoffset_t in size without any special alignment rules may result in a broken Vector layout that reads back with different values.

Example of the flatbuffers:

// sizeof(JustSmallStruct) == 2
// alignof(JustSmallStruct) == 1
struct JustSmallStruct {
  var_0: uint8;
  var_1: uint8;
}

table SmallStructs {
  small_structs: [JustSmallStruct];
}

now if a Vector of just 3 values of JustSmallStruct: { 2, 1 }, { 3, 1 }, { 4, 1 } is created with FlatBufferBuilder::CreateVectorOfStructs<T>() we get this binary Vector that is when read back get zeros in the first JustSmallStruct item:

03 00 00 00 00 00 02 01 03 01 04 01
          ^    ^
          |    |
  vector len   |
               |
         followed by two zero bytes of an alignment
         that was added because uoffset_t required
         4 bytes alignment!?

More confusing is that FlatBufferBuilder::CreateUninitializedVectorOfStructs<T>() method doesn't have this behavior and not adding "2 bytes zeros" before writing the vector len element.

The Fix:
This MR makes builder.CreateVectorOfStructs<T>() behaves correctly for small structures vectors and it's now the same as the FlatBufferBuilder::CreateUninitializedVectorOfStructs<T>().

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[dbaileychess]

Can we add an annotated flatbuffer to this alignment test? (See docs). This would be useful to see how the data is laid out in the binary which is important. Ideally we would check in an annotated binary before this fix, so we can show how it gets fix (sort of like how you did in your PR description).

[bexcite]

@dbaileychess thanks for the feedback, I'll look to the annotated flatbuffer things and will fix the build errors that I see on some compilers/OSs.

[bexcite]

I should admit this alignment_test looks a little bit convoluted, because I decided to extend the original test that essentially broke the alignment ..., maybe it make sense to extract alignment_test related files to a separate folder?

[github-actions]

This pull request is stale because it has been open 3 weeks with no activity. Please comment or label not-stale, or this will be closed in 7 days.

[bexcite]

@dbaileychess I don't think that I can add not-stale label myself, but the PR seems still relevant. Could someone look at it and give a feedback what else I need to do to move it forward? As of now I think that everything was addressed, besides some explicitly left conversations/comments that I put there.

Thanks!

[github-actions]

This pull request is stale because it has been open 3 weeks with no activity. Please comment or label not-stale, or this will be closed in 7 days.

[bexcite]

Added the issue, because I've realized that I posted a fix for the "non-existent" issue so people maybe not looking at it as a bug:

Closes: #8024

[dbaileychess]

Sorry for the delay, was out on paternity leave.

[bexcite]

Hooray! Thanks for accepting and congrats!