Fix Java failure when chunking is combined with uplevels

[robander]

Description

Fixes a bug with maps that use by-topic chunking but also have content outside of the map directory.

The update ensures that when a file is split into multiple chunks, the new chunked topics are created in the same directory as the original topic. Currently, this works correctly when the map is in the root of the temp directory, but fails when the map is not in the root of the temp directory due to uplevels processing.

The fix changes how the temporary file name for the new chunk is created. Currently, the build properly determines what the result file would be in the original source context. However, when it creates the file name used to generate the fragment, it includes the full path from the root of temp, and then appends that to the path of the current topic.

For example, assume I have this construct in the temp dir:

common/reuse.dita
main/input.ditamap
main/topic.dita

When topic.dita is split into multiple chunks, file names are created based on the IDs of each topic. For any but the first-in-file, those file names result in a relative path from temp such as main/id.dita. This is then evaluated against the topic document itself, resulting in an attempt to open main/main/id.dita. This fails, which means variables are not initialized, which results in a build failure later in the chunking step:
Error: java.util.NoSuchElementException

The update ensures we use only the base file name of the new chunk, relative to the document that is splitting.

Motivation and Context

Encountered in customer document that uses the same organization as the new sample integration test doc.

How Has This Been Tested?

Passes unit / integration tests; new integration test added that is fixed with the update.

Type of Changes

• Bug fix (non-breaking change which fixes an issue)

Checklist

• My code follows the code style of this project.
  • https://github.com/dita-ot/dita-ot/wiki/Java-Coding-Conventions
  • https://github.com/dita-ot/dita-ot/wiki/XSLT-Coding-Conventions
• I have updated the unit tests to reflect the changes in my code.

[robander]

This is somewhat related to #2728, which had the same failure, but not related to uplevels.

[robander]

Reworked after discussion with Jarno about how the file name generation was working.

Current version is changed to:

• Generate the file name relative to the file that's being chunked (rather than relative to the map). This means that when file/directory names are relevant, splitting up a document results in new chunks within the same directory
• The temporary chunk name is created relative to the temp directory, as before. To fix the original bug that prompted this, that value is now made absolute by evaluating it relative to the same temp directory, rather than relative to the chunked document.

[jelovirt]

Small clean-up request, otherwise LGTM now.

[jelovirt]

Maybe remove this comment because the log shown here is not what the final modifications will output.

[robander]

Shouldn't have been in there either way, leftover from my debugging of the test file itself. Thanks...