fix: preserve file version snapshots during version creation

[zubeydecivelek]

In _versions(), when we build cumulative file snapshots, we are assigning the same versioned_files dictionary to every version:

versions[version]["files"] = versioned_files

Because this dictionary keeps getting updated in the next loop iterations, the earlier versions were also reflecting the latest state.

For example, if we have:

• draft.pdf v1
• draft.pdf v2

we would expect:

• record version 1 -> draft.pdf v1
• record version 2 -> draft.pdf v2

But because all versions shared the same dictionary reference, we are getting:

• record version 1 -> draft.pdf v2
• record version 2 -> draft.pdf v2

So we are missing previous versions and always keeping the latest file version in the earlier snapshots too.

this one has easy fix:

versions[version]["files"] = deepcopy(versioned_files)

Second issue: we create the version bucket before checking whether the file should be skipped. For example, if Plot files have versions 3 and 4, but plots are skipped later, we still create record versions 3 and 4. Since no kept files change in those versions, they become duplicate versions.

For example:

• main.pdf has 2 version
• plot.png has 4 version

If plots are skipped, we probably only want:

• record version 1 -> main.pdf v1
• record version 2 -> main.pdf v2

But currently we can get:

• record version 1 -> main.pdf v1
• record version 2 -> main.pdf v2
• record version 3 -> same as version 2
• record version 4 -> same as version 2

[zubeydecivelek]

To see the difference here is the dummy PR with the test file only
zubeydecivelek#5